A comprehensive cabinet testing device and a detection method thereof

Abstract

The application discloses a comprehensive cabinet testing device and a detection method thereof, and particularly relates to the technical field of comprehensive cabinet state detection, and is used for solving the problem that the existing technology cannot quantize the gradual drift of the signal monitoring unit under the transient electromagnetic interference, and the device degradation characteristics and measurement errors are difficult to decouple; a high-strength transient electromagnetic interference environment conforming to the physical characteristics of switch operation is simulated in the cabinet; a high-precision reference sensor is connected in parallel at the signal monitoring unit collection end to realize double-track signal millisecond-level synchronous collection; a difference sequence representing drift is generated; an adaptive fault excitation is dynamically generated based on the interference characteristic frequency band of the difference sequence; the mixed signal is subjected to covariance matrix orthogonal decomposition to separate the drift characteristic vector and the fault characteristic vector and calculate the projection scalar value; finally, the historical reference output drift interference coefficient and the three-level fault confidence report are combined to significantly improve the reliability of the predictive maintenance of the intelligent power distribution system.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of integrated cabinet state detection, and more particularly to an integrated cabinet testing device and a detection method thereof. BACKGROUND

[0002] The intelligent integrated cabinet for power distribution system (such as intelligent circuit breaker cabinet and ring network cabinet) is a key facility for realizing power distribution and control, which integrates switch control devices and signal monitoring units inside for real-time acquisition of electrical parameters. The high-intensity transient electromagnetic interference generated by the action of large current switches in the operation of such cabinets is a core factor affecting the measurement accuracy of the monitoring units. The existing technology verifies the anti-interference performance of the cabinet system through standardized electromagnetic compatibility (EMC) testing to ensure that the monitoring function does not fail.

[0003] The existing detection of integrated cabinets only focuses on the functional integrity of the monitoring units, and cannot quantitatively evaluate the gradual drift of the micro signals generated under continuous electromagnetic stress, resulting in difficulty in effectively decoupling the real device degradation characteristics and signal measurement errors in the health assessment model of the cabinet system, which reduces the confidence of the predictive maintenance data of the cabinet and directly affects the safe operation of the intelligent power distribution system. SUMMARY

[0004] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present application provide an integrated cabinet testing device and a detection method thereof to solve the problems raised in the background art.

[0005] To achieve the above-mentioned purpose, the present application provides the following technical solutions:

[0006] An integrated cabinet detection method, comprising:

[0007] S1, simulating a high-intensity transient electromagnetic interference environment in the cabinet;

[0008] S2, connecting a reference sensor in parallel to the signal acquisition end of the signal monitoring unit, and synchronously collecting the output data of the reference sensor and the original analog signal of the signal monitoring unit;

[0009] S3, calculating the difference between the original analog signal of the signal monitoring unit and the output data of the reference sensor at the same time point to generate a difference sequence;

[0010] S4, extracting the feature frequency band with concentrated interference energy in the difference sequence as an injection frequency reference, applying a preset fault excitation according to the injection frequency reference, and synchronously collecting the output data of the reference sensor and the original analog signal of the signal monitoring unit to generate a fault feature data set;

[0011] S5, performing covariance matrix-based orthogonal decomposition on the difference sequence and the fault feature dataset to separate a drift feature vector and a fault feature vector, and calculating a projection scalar value of the drift feature vector in the direction of the fault feature vector;

[0012] S6, outputting a drift interference coefficient and a fault confidence report based on the drift feature vector, the fault feature vector, and the projection scalar value.

[0013] In a preferred embodiment, a high-intensity transient electromagnetic interference environment is simulated in the cabinet, including:

[0014] A surge generator is used to inject a current pulse into the power loop of the switch control device in the cabinet, the rise time of the current pulse corresponds to the natural response time range of the large current switch action, the duration corresponds to the transient decay period caused by switch operation, and the current pulse spectrum characteristics match the typical switch operation spectrum defined in the standard interference waveform library.

[0015] In a preferred embodiment, a reference sensor is connected in parallel to the signal to be measured of the signal monitoring unit, and the reference sensor output data and the original analog signal of the signal monitoring unit are synchronously collected, including:

[0016] The reference sensor signal input terminal is directly electrically connected in parallel to the signal to be measured collection line of the signal monitoring unit through a conductive connecting piece;

[0017] The reference sensor output data and the original analog signal of the signal monitoring unit are synchronously collected by using a double-channel synchronous collection device:

[0018] One channel of the double-channel collection device is connected to the reference sensor output terminal for collecting the reference sensor output data, and the other channel is connected to the analog signal terminal of the signal monitoring unit for collecting the original analog signal of the signal monitoring unit;

[0019] The transmission lines from the reference sensor output terminal to the collection device channel and from the signal monitoring unit analog signal terminal to the collection device channel are connected by shielded twisted pair lines, and the length difference of the two cable lines is controlled within a negligible range;

[0020] The collection start command is initialized to the two collection channels by an external trigger pulse.

[0021] In a preferred embodiment, the difference between the original analog signal of the signal monitoring unit and the reference sensor output data at the same time point is calculated to generate a difference sequence, including:

[0022] The data sequences of the reference sensor output data and the original analog signal of the signal monitoring unit stored by the double-channel synchronous collection device are obtained;

[0023] The time domain alignment processing of the data sequence of the reference sensor output data and the original analog signal of the signal monitoring unit adopts the same timestamp point matching mechanism;

[0024] The difference value operation is performed at the matched timestamp point, and the operation objects are the instantaneous values of the original analog signal of the signal monitoring unit and the instantaneous values of the reference sensor output data;

[0025] The difference value operation result is sequentially stored with the timestamp point as the independent variable to generate a difference value sequence;

[0026] The difference value sequence is transmitted to the cache area in real time before performing zero mean preprocessing to suppress the direct current component drift interference.

[0027] In a preferred embodiment, the feature frequency band with concentrated interference energy in the difference value sequence is extracted as an injection frequency reference, including:

[0028] The Fourier transform is performed on the difference value sequence to obtain the frequency spectrum distribution data, and the frequency band interval with a frequency spectrum amplitude exceeding ten times the average energy level is identified as the feature frequency band with concentrated interference energy;

[0029] The center frequency of the feature frequency band is set as the injection frequency reference, and a preset fault excitation signal centered on the corresponding injection frequency reference is generated through a programmable signal generator.

[0030] In a preferred embodiment, the preset fault excitation is applied according to the injection frequency reference, and the reference sensor output data and the original analog signal of the signal monitoring unit are synchronously collected to generate a fault feature data set, including:

[0031] The preset fault excitation signal is connected to the signal injection port of the switch control device in the cabinet through a power amplifier;

[0032] During the application of the preset fault excitation, the collection operations of the reference sensor output data and the original analog signal of the signal monitoring unit are synchronously performed, and the collection time window covers the complete excitation period;

[0033] The collected reference sensor output data and the original analog signal of the signal monitoring unit are timestamp-aligned and packaged to generate a fault feature data set.

[0034] In a preferred embodiment, the difference value sequence and the fault feature data set are subjected to orthogonal decomposition based on the covariance matrix to separate the drift feature vector and the fault feature vector, and the projection scalar value of the drift feature vector in the direction of the fault feature vector is calculated, including:

[0035] A joint sample matrix of the difference value sequence data points and the fault feature data set data points is constructed, and a two-dimensional data array is formed according to the same timestamp points;

[0036] A covariance matrix of the two-dimensional data array is calculated, and a covariance matrix element value represents a statistical correlation degree between the difference value sequence data and the fault feature data;

[0037] An eigenvalue decomposition operation is performed on the covariance matrix to obtain a sequence of eigenvectors arranged in descending order of eigenvalue size;

[0038] An eigenvector corresponding to the largest eigenvalue is defined as a fault feature vector, and an eigenvector corresponding to the second largest eigenvalue is defined as a drift feature vector;

[0039] A projection scalar value is calculated by a dot product operation of the drift feature vector and the fault feature vector, and the dot product operation result represents a projection length of the drift feature vector in the direction of the fault feature vector.

[0040] In a preferred embodiment, a drift interference coefficient and a fault confidence report are output based on the drift feature vector, the fault feature vector and the projection scalar value, including:

[0041] The drift interference coefficient is defined by an exponential function processing result of the projection scalar value, the base number of the exponential function is set to a natural constant, and the exponential term is a negative proportional amplification value of the projection scalar value;

[0042] A fault confidence reference value is calculated according to a square of a two-norm of the fault feature vector, and the square of the two-norm represents an energy accumulation order of magnitude of the fault feature vector;

[0043] A relative confidence proportion is obtained by comparing the fault confidence reference value with a preset historical health operation period reference value;

[0044] A fault confidence level label is marked according to an interval mapping relationship between the relative confidence proportion and a preset threshold value;

[0045] The drift interference coefficient and the fault confidence level label are combined to generate a structured drift interference coefficient and fault confidence report.

[0046] In another aspect, the application provides a comprehensive cabinet testing device, comprising:

[0047] An interference simulation module simulates a high-intensity transient electromagnetic interference environment in the cabinet;

[0048] A signal parallel module connects a reference sensor in parallel at a to-be-tested signal acquisition end of the signal monitoring unit to synchronously acquire reference sensor output data and original analog signals of the signal monitoring unit;

[0049] A sequence generation module calculates a difference value between the original analog signals of the signal monitoring unit and the reference sensor output data at the same time point to generate a difference value sequence;

[0050] Feature injection module: extract the feature frequency band of the interference energy set in the difference sequence as the injection frequency reference, apply the preset fault excitation according to the injection frequency reference, and synchronously collect the output data of the reference sensor and the original analog signal of the signal monitoring unit to generate the fault feature data set;

[0051] Vector decoupling module: orthogonal decomposition based on the covariance matrix is performed on the difference sequence and the fault feature data set to separate the drift feature vector and the fault feature vector, and the projection scalar value of the drift feature vector in the direction of the fault feature vector is calculated;

[0052] Evaluation output module: based on the drift feature vector, the fault feature vector and the projection scalar value, the drift interference coefficient and the fault confidence report are output.

[0053] Compared with the prior art, the present application has the following beneficial effects:

[0054] 1. By accurately reproducing the transient electromagnetic environment of the switch operation physical characteristics in the cabinet, and introducing a high-precision reference sensor in parallel with the monitoring unit, a double-track synchronous signal acquisition system is constructed, a true value reference benchmark is established directly from the signal source, the drift amount of the original analog signal of the signal monitoring unit is captured under strong interference background, and the difference sequence generation mechanism based on time domain alignment converts the traditional difficult-to-catch small progressive drift into quantifiable time series data stream, solving the pain point that the existing technology cannot distinguish between real device degradation and signal measurement error.

[0055] 2. By dynamically binding the interference feature frequency band of the difference sequence and the fault excitation parameter, the fault injection process is adaptively matched with the actual drift characteristics of the cabinet, forming an active detection mechanism of the drift-fault coupling relationship; the mixed signal is decoupled into drift feature vectors and fault feature vectors with clear physical meaning by using the orthogonal decomposition technology of the covariance matrix, and then the coupling strength of the two is accurately quantified by the projection scalar; finally, combined with the confidence dynamic evaluation system established based on the historical operation reference, the drift interference coefficient and the three-level confidence label are output, providing a directly executable decision basis for predictive maintenance, and significantly improving the reliability and operability of the device state evaluation. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1 The flowchart of the comprehensive cabinet detection method of the present application;

[0057] Figure 2 The structural schematic diagram of the comprehensive cabinet test device of the present application. DETAILED DESCRIPTION

[0058] With reference to the accompanying drawings, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all the other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0059] Embodiment 1 Figure 1 A comprehensive cabinet detection method is given, comprising:

[0060] S1, simulating a high-intensity transient electromagnetic interference environment in the cabinet;

[0061] S2, connecting a reference sensor in parallel to a signal monitoring unit at a signal to be measured collection end, and synchronously collecting reference sensor output data and signal monitoring unit original analog signals;

[0062] S3, calculating a difference value of the signal monitoring unit original analog signals and the reference sensor output data at the same time point, and generating a difference value sequence;

[0063] S4, extracting a feature frequency band of the interference energy set in the difference value sequence as an injection frequency reference, applying a preset fault excitation according to the injection frequency reference, and synchronously collecting reference sensor output data and signal monitoring unit original analog signals to generate a fault feature data set;

[0064] S5, performing orthogonal decomposition based on a covariance matrix on the difference value sequence and the fault feature data set, separating a drift feature vector and a fault feature vector, and calculating a projection scalar value of the drift feature vector in the direction of the fault feature vector;

[0065] S6, outputting a drift interference coefficient and a fault confidence report based on the drift feature vector, the fault feature vector and the projection scalar value.

[0066] S1, simulating a high-intensity transient electromagnetic interference environment in the cabinet, and the specific implementation is:

[0067] A surge generator with programmable pulse output capability is used to inject current pulses into the power circuit of the switch control equipment in the cabinet to simulate a high-intensity transient electromagnetic interference environment in the cabinet. The surge generator needs to meet the technical specifications of a minimum rise time of not more than 5 ns and a peak current output capability of not less than 10 kA. The setting method of the current pulse rise time is: through a high-speed camera system, the switch contact movement process is captured with a frame interval of less than 10 μs, and a current sensor with a transient response time of less than 10 ns is used to record the time mark point synchronously; the time interval between the contact separation time and the current zero time is defined as the natural response time boundary value. Mapping relationship is established for different switch media types: in vacuum environment, the boundary value is calculated by ionized particle diffusion rate model, and in gas environment, the boundary value is determined based on molecular ionization recombination rate, forming a mapping database of switch type and time boundary value for calling.

[0068] The setting method of the current pulse duration is: a current sensing device with a bandwidth covering 200 MHz is deployed in the switch power circuit, and the opening and closing operation waveform data is collected. The peak value point position of the current waveform is identified, and the envelope line equation is established by using the polynomial least square fitting for the decay curve after the peak value point. The time domain coordinate point corresponding to the current value from the peak value to 1% of the steady value is solved, and the time difference between this point and the peak value point is the transient decay period. In order to eliminate the randomness of measurement, the median value of the time difference is taken as the reference value after repeating the experiment 10 times.

[0069] The matching method of current pulse spectrum characteristics is realized through the following process: a database containing switch operation spectrum reference templates is established. When performing matching, the surge generator output signal is connected to the spectrum acquisition device through a 20 dB attenuation device; the resolution bandwidth of the spectrum acquisition device is set to be not more than 1 kHz; the frequency range is scanned to cover 0 Hz to 100 MHz; the frequency domain similarity index of the measured spectrum and the target template is calculated in real time, and the calculation method is the ratio of the dot product of the measured spectrum amplitude sequence and the template amplitude sequence to the product of the two norm products; when the similarity index exceeds 90%, it is determined that the matching is passed. The setting basis of the similarity index threshold value of 90% is that below this value, the pulse energy distribution error will exceed the scheme tolerance. If the matching fails, the compensation mechanism is triggered according to the similarity index interval: when the index is between 80% and 90%, the resistance value of the inductance element in the pulse shaping network is increased, and the compensation step is 1% of the initial value; when the index is less than 80%, the spectrum template is reselected.

[0070] The performance constraints that the execution device needs to meet include: surge generator rise time ≤ 5ns; high-speed camera system time resolution ≤ 10μs; spectrum acquisition device dynamic range > 80dB. The abnormal working condition processing logic covers the unknown scene of the switch type: traverse all template categories in the database, select the template corresponding to the highest similarity index; when there is non-design band interference, insert a filter with a center frequency aligned with the interference frequency point and a stop band bandwidth of 2 times the interference bandwidth into the output path.

[0071] S2, parallel reference sensors at the signal monitoring unit's signal acquisition end, synchronously collect reference sensor output data and signal monitoring unit original analog signals, and the specific implementation is as follows:

[0072] The specific operation of connecting reference sensors at the signal monitoring unit's signal acquisition end is as follows: a copper conductive connecting piece is used as a transition component to directly electrically connect the reference sensor signal input terminal and the signal monitoring unit signal acquisition line in parallel. Before connection, the resistivity of the metal material of the reference sensor input terminal and the signal monitoring unit acquisition line is measured, and a copper sheet with a resistivity matching degree of more than 95% is selected as the connecting piece. When implemented, a professional crimping tool is used to crimp the connecting piece at both ends of the reference sensor input terminal and the exposed conductor part of the signal monitoring unit acquisition line, respectively, and the crimping pressure is controlled in the range of 3kN to 5kN to ensure that the micro deformation of the contact interface is sufficient. After connection, the contact resistance is measured by a four-wire Kelvin bridge, and the resistance value is controlled to be not more than 0.2mΩ. After completion of the connection, the temperature rise change of the contact area is observed, and the temperature rise is not more than 5K after continuous operation for 1 hour, which is considered qualified.

[0073] The actual operation process of synchronously collecting reference sensor output data and signal monitoring unit original analog signals is as follows: both channels of the dual-channel synchronous acquisition device use high-precision analog-to-digital converters to realize signal acquisition, and the input impedance of each channel is set to 10MΩ. The channel for collecting reference sensor output data is connected to the reference sensor output terminal through a low-noise coaxial cable, and the end of the cable is fixed to the terminal connection post by right-angle welding process; the channel for collecting signal monitoring unit original analog signals is connected to the signal monitoring unit analog signal terminal through a cable of the same specification, and the terminal contact surface is pre-polished to remove the oxidation layer. After connection, the loop impedance is verified using an impedance analyzer, and the resistance of each loop is required to be less than 5mΩ. Under the condition of signal amplitude 1V, the consistency of the channel path gain is tested, and the deviation is controlled within 0.1%.

[0074] The implementation scheme of clock offset control when the dual-channel acquisition clock is driven by the same crystal oscillator is as follows: the crystal oscillator is selected to be a constant temperature control type quartz crystal oscillator, and the frequency stability is better than 1x10 -9The crystal oscillator output end is divided into two paths through impedance matching network, and is connected to the clock distribution network of the two acquisition channels. The clock distribution network adopts tree-shaped symmetric topology, and the signal transmission delay of each buffer is controlled within 50 ps. The key of clock offset control lies in dynamic calibration mechanism: before each acquisition start, a step test signal is injected to measure the actual sampling time difference of the two channels, and when the value exceeds 500 ps, the compensation process is started. The compensation operation is realized by adjusting the adjustable capacitor elements in the clock distribution network, so that the time difference falls within the target value of 200 ps. For example, for the case of 1 MS / s sampling rate, the target time difference is controlled within 1 ns, which is one thousandth of the sampling interval.

[0075] The transmission lines from the reference sensor output terminal to the acquisition device channel and from the signal monitoring unit analog signal terminal to the acquisition device channel are connected by shielded twisted pair as follows: first, high-quality twisted pair cable with shield coverage of more than 95% is selected, and the cable diameter is controlled within 2.8 mm ± 0.1 mm. The two transmission lines are cut according to the principle of equal length, and the length measurement uses laser ranging equipment with an accuracy of 0.1 mm. When wiring, the two cables are laid in parallel, with a spacing of twice the cable diameter, and the bending radius is greater than 15 times the wire diameter. Double shielding grounding is implemented at both ends of the cable: the near-end grounding line is directly connected to the cabinet grounding row, and the far-end is suppressed by a ferrite magnetic ring to suppress high-frequency circulating current. All connectors are selected with gold-plated copper core structure, and the connection points are coated with conductive silicone grease to enhance high-frequency conduction.

[0076] The technical support means for controlling the length difference of the two cables within a negligible range includes two aspects: in the design stage, electromagnetic field simulation software is used to calculate the signal delay difference caused by the length difference of the cable, and the path planning is optimized to make the theoretical delay difference less than 100 ps; in the construction stage, an optical calibration device is used to monitor the cable laying condition in real time, and when the length difference exceeds 0.5 mm, an alarm is triggered to adjust. After completion of laying, the actual electrical length difference is measured by time domain reflectometer, and the signal propagation time difference caused by the difference is controlled to be less than 1 ns.

[0077] The acquisition start command is synchronized to the operation process of the two acquisition channels by an external trigger pulse as follows: the pulse generator outputs a trigger signal with a pulse width of 100 ns, which is transmitted to the splitter through a 50 Ω coaxial cable. The splitter adopts Wilkinson power divider structure, and the power division imbalance is less than 0.1 dB in the frequency range of 1 MHz to 100 MHz. The distributed trigger signal is connected to the trigger input interface of the two acquisition channels through equal-length cables. In order to ensure the accurate synchronization of the trigger time, the clock source of the pulse generator and the crystal oscillator source of the acquisition device are phase-locked through a 10 MHz reference clock.

[0078] The abnormality processing mechanism covers three typical scenarios: when the signal amplitude exceeds the range, the acquisition device automatically switches the range and marks the overrun point; when the clock drift between channels is detected to be continuously increasing, the crystal oscillator temperature control module is started to fine-tune the frequency compensation; when the impedance mismatch of the transmission line causes the reflection to exceed the threshold, the rewiring process is triggered. All abnormal data are recorded with timestamps and fault codes for subsequent analysis, and the fault scenario is reproduced by the calibrated test equipment for regression verification.

[0079] S3, calculate the difference between the original analog signal of the signal monitoring unit at the same time point and the output data of the reference sensor, generate a difference sequence, and the specific implementation is as follows:

[0080] The specific operation of obtaining the data sequence of the reference sensor output data and the original analog signal of the signal monitoring unit stored by the dual-channel synchronous acquisition device is as follows: the storage in the dual-channel synchronous acquisition device is configured as a dual-channel independent storage area, and each channel is allocated a dedicated physical address space. The starting address of the reference sensor output data storage area is fixedly allocated at the 0 address block, and the starting address of the signal monitoring unit original analog signal storage area is set at a position 8MB away from the reference area. The data sequence is stored in a fixed interval manner, and each sampling point record contains timestamp information and value information. The timestamp information is stored in the low 2 bytes of space, and the value information is stored in the high 2 bytes of space. The representation range of the sampling point value is determined according to the bit width of the analog-to-digital converter, for example, for a system with 16-bit sampling accuracy, the value range is from -32768 to +32767.

[0081] The operation method of performing time domain alignment processing on the data sequence of the reference sensor output data and the original analog signal of the signal monitoring unit is as follows: read the timestamp information from the storage area, and the timestamp accuracy reaches 1ns. The frequency multiplication phase-locked technology is used to realize the accuracy guarantee of 1000 times higher than the sampling time interval. The alignment processing process is divided into three steps: the first step is to select the reference time point in the reference sensor data sequence; the second step is to search for the sampling point with the smallest time difference with the reference point in the signal monitoring unit data sequence; the third step is to establish a matching pair relationship and mark the time offset compensation value. When the time deviation exceeds the preset threshold of 10ns, a linear interpolation algorithm is used to generate a virtual matching point. The calculation error of the entire matching process is controlled within 1% of the sampling interval.

[0082] The technical scheme of difference operation at the time stamp point of completed matching is as follows: the instantaneous value of the reference sensor output data of completed matching and the instantaneous value of the original analog signal of the signal monitoring unit are extracted, and both values are signed integer formats. The difference operation strictly performs the mathematical operation of subtracting the instantaneous value of the reference sensor output data from the instantaneous value of the original analog signal of the signal monitoring unit, and the operation result is temporarily stored in the form of a 16-bit signed integer. When the operation result exceeds the range of ±32767, it is automatically switched to a 32-bit long integer storage format. The sequence index value is updated after each operation step to ensure real-time recording of the processing progress.

[0083] The implementation steps of sequentially storing the difference operation result to generate a difference sequence include: allocating an independent cache area for the difference sequence, and using a continuous physical address layout for the cache area address space. Each difference sequence record includes two fields of time difference and operation result, the time difference field stores the offset of the current time stamp from the sequence starting time, and the operation result field stores the difference calculation result. The storage process is automatically executed through a direct memory access controller, and the address pointer is automatically increased by 4 bytes after processing each record. The storage format uniformly adopts the IEEE 754 single-precision floating-point binary encoding standard, and the integer operation result is automatically converted to floating-point format when stored, and the conversion method is to retain the original integer value and add a zero-value decimal part.

[0084] The processing flow of performing zero mean preprocessing before the difference sequence is transmitted to the cache area is as follows: first, calculate the 512-point moving average of the current difference sequence. The preprocessing operation is defined as the difference between the current point difference and the moving average. The update frequency of the moving average is set to be recalculated every 10 new sampling points processed. The verification method of the preprocessing effect is to calculate the variance change rate of the sequence before and after processing, for example, the variance is required to be reduced by more than 50%. The sequence after preprocessing is transmitted to the next cache area through a 64-bit bus.

[0085] The abnormal processing mechanism covers three cases: when the time stamp matching failure rate is continuously >5%, the time sequence interpolation compensation algorithm is started; when 10 consecutive difference operation overflows are detected, the gain setting is automatically adjusted; when the remaining space of the cache area is <1KB, the lossless compression algorithm is executed. All processing logics are implemented in a 5-stage pipeline architecture in a programmable logic device with a working frequency of 200MHz.

[0086] S4, extract the feature frequency band of the interference energy set in the difference sequence as the injection frequency reference, apply a preset fault excitation according to the injection frequency reference, and synchronously collect the reference sensor output data and the original analog signal of the signal monitoring unit to generate a fault feature data set, which is specifically implemented as:

[0087] The implementation process of the Fourier transform on the difference sequence to obtain the spectrum distribution data is as follows: the complete data sequence is read from the difference sequence storage buffer generated in S3. The Fourier transform is performed by using a 2048-point fast Fourier transform algorithm, and the data sequence is multiplied by a Hanning window function before the transform to suppress spectrum leakage. The spectrum distribution data calculation process is as follows: first, the complex Fourier transform is performed, then the power spectrum density value is obtained by taking the modulus of the transform result of each frequency point and squaring, and finally the power spectrum density is stored as a double-precision floating point array. The frequency resolution of the power spectrum array is determined according to the sampling frequency and the number of transform points, for example, the resolution of 2048-point transform is about 48.8Hz when the sampling frequency is 100kHz.

[0088] The technical implementation of identifying the frequency band interval with a spectrum amplitude exceeding ten times the average energy level as the characteristic frequency band of the concentrated interference energy includes three steps: the first step is to calculate the arithmetic mean of the power spectrum array as the average energy level reference; the second step is to scan the array to find a frequency band with more than twenty continuous frequency points and a power value continuously greater than ten times the average energy level; and the third step is to select the frequency band with the largest power integral value from the candidate frequency band as the characteristic frequency band. The characteristic frequency band boundary determination method is: the starting frequency is the first frequency point of the continuous super-threshold frequency band, the ending frequency is the last super-threshold point, and the center frequency is calculated as (starting frequency + ending frequency) ÷ 2. For example, if it is found that the power values of all points in the 350kHz to 450kHz frequency band are >10 times the average value, the center frequency is set to 400kHz.

[0089] The operation procedure for setting the center frequency of the characteristic frequency band as the injection frequency reference is as follows: the center frequency value is written into the frequency control register in 32-bit floating point format. When the center frequency is lower than 10kHz, it is automatically corrected to 10kHz to ensure detectability; when the center frequency exceeds 1MHz, it is limited to 1MHz to prevent high-frequency interference. The frequency writing operation is realized through direct register mapping at the hardware level, with a delay time of not more than 100ns. The injection frequency reference value is also backed up to the non-volatile memory for audit traceability.

[0090] The technical solution for generating a preset fault excitation signal centered on the corresponding injection frequency reference by a programmable signal generator includes: the programmable signal generator receives the injection frequency reference value through the SPI interface. The fault excitation type setting rule is: for switch contact fault, a ramp waveform with a frequency equal to the injection frequency reference value, a starting amplitude of 30% of the rated value, and an increment of 2% per millisecond is generated; for insulation deterioration fault, an amplitude modulation wave with the injection frequency reference value as the carrier frequency and 80% depth sinusoidal modulation is generated. The signal parameters are stored in an XML configuration file, for example, the terminal amplitude of the ramp waveform is set to 70% of the rated value. The signal verification method is: the output end is connected to an oscilloscope to measure the actual frequency, and the requirement is that the deviation between the set frequency and the actual frequency is ≤0.1%.

[0091] The connection specification of the preset fault excitation signal through the power amplifier to the signal injection port of the switch control device in the cabinet requires that the power amplifier gain is set to 20 dB fixed gain, and the input / output impedance matching is 50 Ω ± 5%. The excitation signal output end is connected to the standby control terminal of the switch control device through a shielded twisted pair, and a crimp terminal is used at the connection point to ensure that the contact resistance is less than 0.5 mΩ. The safety protection measures include a 0.5 A fuse in series and a bidirectional transient suppression diode in parallel at the output end. The test return loss is less than or equal to -30 dB after connection, which is considered qualified.

[0092] The implementation details of the synchronous execution of the reference sensor output data and the signal monitoring unit original analog signal acquisition operation during the application of the preset fault excitation are as follows: the acquisition start trigger signal is set 200 μs after the rising edge of the excitation signal to ensure signal stability. The acquisition time window is dynamically set according to the excitation period: for an excitation signal with a period ≤ 50 μs, the acquisition time window is set to 500 μs, and for an excitation signal with a period > 50 μs, three complete periods are acquired. The double-channel acquisition uses a hardware synchronization trigger mechanism, and the trigger jitter is controlled within 500 ps. For example, the acquisition time window of a 400 kHz excitation signal (period 2.5 μs) is fixed at 500 μs.

[0093] The specific process of packaging the collected reference sensor output data and signal monitoring unit original analog signals according to the timestamp to generate the fault feature data set is as follows: first, read the original data files from the two acquisition channels respectively; second, align the time axis using a linear interpolation algorithm, with a timestamp tolerance of one ten-thousandth of the sampling interval; third, fill in the mismatched points using cubic spline interpolation. The packaging format uses fixed 8 bytes per point: 4 bytes of timestamp (μs level precision) + 2 bytes of reference sensor data + 2 bytes of monitoring unit data. The file structure includes a 128-byte file header (records excitation parameters) and a continuous data block, for example, 10,240 samples generate an 80 KB file.

[0094] The abnormality handling mechanism is divided into three levels: when the excitation signal distortion degree > 5%, the output is automatically cut off and the excitation is re-initiated; when no qualified feature band is detected, the historical average frequency is used as the injection frequency reference; when the acquisition time synchronization error > 1 μs, the software timestamp correction algorithm is enabled. All operation logs record the complete transaction process with a time tag accuracy of 10 ms. The technical effect verification method is to inject a standard signal with a frequency error < 0.01% into the system, verify the feature band recognition accuracy error ≤ ± 0.5%, the excitation signal frequency error ≤ ± 0.1%, and the time synchronization error ≤ 10 ns.

[0095] S5, orthogonal decomposition based on covariance matrix is performed on the difference sequence and the fault feature data set to separate the drift feature vector and the fault feature vector, and the projection scalar value of the drift feature vector in the direction of the fault feature vector is calculated, which is specifically implemented as:

[0096] The technical implementation details of constructing the joint sample matrix of the difference sequence data points and the fault feature dataset data points are as follows: all data point samples are read from the difference sequence storage area generated in step S3, and corresponding data samples are read from the fault feature dataset storage area generated in step S4. The time matching algorithm used in the same timestamp alignment processing is executed: the timestamp value is extracted from the difference sequence, and the time point with a difference absolute value not exceeding 0.001% of the sampling time interval is searched in the timestamp sequence of the fault feature dataset for matching. When the matching time deviation exceeds the threshold, a virtual sample point is generated by performing linear interpolation on the fault feature dataset side. The construction format of the two-dimensional data array is fixed as an N-row x 2-column matrix, where N is equal to the total number of matching successful time points, each time point occupies a row, the first column stores the difference sequence data value, and the second column stores the average value of the reference sensor output data and the original analog signal of the signal monitoring unit at the corresponding time in the fault feature dataset. For example, when the difference sequence contains 32768 points, the 32768 points of the fault feature dataset form a 32768 x 2 matrix.

[0097] The technical implementation process of calculating the covariance matrix of the two-dimensional data array includes three core steps: first, calculate the arithmetic mean of the first column (difference sequence column) of the two-dimensional data array as μ1, and the arithmetic mean of the second column (feature data column) as μ2; then, center the elements in the matrix: subtract μ1 from the first column element value to obtain the centered difference vector, and subtract μ2 from the second column element value to obtain the centered feature vector; finally, the covariance matrix element calculation rule is: the first row and first column element of the matrix is equal to the autocovariance value of the centered difference vector, the second row and second column element of the matrix is equal to the autocovariance value of the centered feature vector, and the first row and second column element of the matrix is equal to the cross-covariance value of the centered difference vector and the centered feature vector. The element value of the covariance matrix is in the square of the original data unit, for example, the voltage data unit is V 2 . After the covariance matrix is calculated, the matrix condition number detection is performed, and when the condition number is greater than 1000, the data standardization preprocessing is automatically triggered.

[0098] The detailed method of performing eigenvalue decomposition operation on the covariance matrix is as follows: the Jacobi rotation iterative algorithm is used for eigenvalue decomposition, and the iterative process includes initializing the rotation matrix as the unit matrix, calculating the maximum value position (i, j) of the non-diagonal element, constructing the rotation matrix to zero the (i, j) element, and setting the iteration termination condition as all non-diagonal element absolute values being less than 10 -10The output of the eigenvector sequence is sorted in descending order of eigenvalues, with the eigenvector corresponding to the largest eigenvalue as the first element of the sequence, and each eigenvector stored as a double-precision floating-point number array. The operation is completed on a processor equipped with a 128-bit floating-point operation unit, and the upper limit of the single decomposition time is 50 ms. For example, when processing a 1000x1000 covariance matrix, 100 iterations are required.

[0099] The eigenvector corresponding to the largest eigenvalue is defined as the fault eigenvector, and the eigenvector corresponding to the second largest eigenvalue is defined as the drift eigenvector. The implementation rules are as follows: extract the eigenvector with index value 0 from the eigenvector sequence as the fault eigenvector, and extract the eigenvector with index value 1 as the drift eigenvector. The validity verification of the eigenvector includes two dimensions: one is the vector norm test, which calculates the square root of the sum of the squares of the components of the vector, and when the deviation from 1 exceeds 0.01, normalization processing is performed; the second is the physical meaning verification, the absolute value of the first component of the fault eigenvector must be greater than 0.5, and the absolute value of the second component of the drift eigenvector must be greater than 0.3, otherwise the covariance matrix calculation process is rechecked. The direction of the fault eigenvector is defined as the first component being positive, and the direction of the drift eigenvector is defined as the second component being positive.

[0100] The technical scheme for calculating the projection scalar value by the dot product operation of the drift eigenvector and the fault eigenvector is as follows: the dot product calculation formula is to multiply each component of the drift eigenvector by the corresponding component of the fault eigenvector and then accumulate the sum. The projection scalar value represents the projection length of the drift eigenvector in the direction of the fault eigenvector, and the value range is determined by the angle between the two vectors, with a value range of [-1, 1]. The calculation result is stored in a double-precision floating-point variable, with six digits after the decimal point. The verification process includes: calculating the product of the two vector norms to obtain the original dot product value, and simultaneously calculating the projection length by the coordinate geometry formula, and when the difference between the two methods exceeds 0.001%, the data review program is started. For example, in a preset scenario where the angle between the two vectors is 60°, the projection scalar value should be 0.500000.

[0101] The abnormal processing mechanism covers three cases: when the condition number of the covariance matrix is too large, a regularization coefficient of the order of 1% is automatically added; when the ratio of the largest eigenvalue to the second largest eigenvalue is less than 2, the eigenvector is marked as unreliable and a warning is issued; when the projection scalar value exceeds the range of [-1.5, 1.5], it is forced to be limited to the effective interval. All calculation operations are performed on a computing platform equipped with 8GB of memory and a four-core processor, and parallel acceleration is achieved using the OpenMP multi-threaded library, with four threads set for synchronous operation.

[0102] S6, based on the drift eigenvector, the fault eigenvector and the projection scalar value, output the drift interference coefficient and the fault confidence report, the specific implementation is:

[0103] The specific technical implementation of the drift interference coefficient defined by the exponential function processing result of the projection scalar value is as follows: receiving the projection scalar value input passed in step S5, which is a double-precision floating-point number ranging from [-1.00, +1.00]. The drift interference coefficient calculation is implemented using an exponential function with the natural constant e as the base, and its power index expression is the negative absolute proportional amplification coefficient multiplied by the projection scalar value. The standard configuration value of the absolute proportional amplification coefficient is 3.0, but when the projection scalar value is detected as negative, it is automatically switched to 5.0 to enhance the response sensitivity of negative drift. The calculation process is completed in a hardware floating-point operation unit, and the execution cycle is controlled within 50ns. Taking a projection scalar value of 0.50 as an example, the drift interference coefficient calculation result is about 0.223. The drift interference coefficient, as a dimensionless proportional parameter, is finally normalized to the range [0.00, 1.00].

[0104] The operation specification for calculating the fault confidence reference value according to the square of the two-norm of the fault feature vector is as follows: read the fault feature vector array from the data buffer area, and the array length is fixed to three elements. The calculation execution flow of the square of the two-norm is: perform a square operation on each component element in the array, and then sum the three square values. This calculation result represents the energy accumulation order of magnitude of the fault feature vector, with the unit being the square of the original physical quantity. The calculation is implemented using an accumulator hardware, and the single calculation delay does not exceed 20ns. For example, when the input vector is (0.30, 0.40, 0.50), the square of the two-norm calculation result is exactly 0.500000. The calculation result is stored in a double-precision floating-point format, with the effective bits reserved to six digits after the decimal point.

[0105] The technical scheme for comparing the fault confidence reference value with the preset historical health operation period reference value includes: extracting the fault confidence reference value data samples of the same type of equipment during the last 30 days of health operation from the historical database. The calculation method of the historical reference value is the arithmetic mean of all sample points in this period, and the sample point collection frequency is once every minute. The calculation formula of the relative confidence ratio is defined as: (current fault confidence reference value - historical reference value) / historical reference value x 100%. The update mechanism of the historical reference value is set to automatically recalculate the average value every time the system accumulates 100 new health sample points. The comparison result is divided into three levels: the relative confidence ratio < 5% is determined as the low-risk interval, ≥5% and <20% is the medium-risk interval, and ≥20% is the high-risk interval.

[0106] The implementation rule for marking the fault confidence level label according to the interval mapping relationship between the relative confidence proportion and the preset threshold value is that the preset three-level determination threshold boundaries are 5% and 20% respectively. The label generation rule is as follows: when the relative confidence proportion is located in the interval [0%, 5%), an "A-level high confidence" label is generated; when the relative confidence proportion is located in the interval [5%, 20%), a "B-level medium confidence" label is generated; and when the relative confidence proportion is located in the interval [20%, 100%], a "C-level low confidence" label is generated. The label data is stored in a single-byte encoding format: the A level corresponds to the ASCII value 65 (character 'A'), the B level corresponds to 66 ('B'), and the C level corresponds to 67 ('C'). The label mapping operation is completed by querying the configuration table pre-stored in the EEPROM.

[0107] The process of combining the drift interference coefficient and the fault confidence level label to generate a structured drift interference coefficient and fault confidence report includes: defining the total length of the report data structure as 64 bytes, which is divided into a 16-byte header information area and a 48-byte data area. The header information area includes an 8-byte timestamp (millisecond level accuracy) and a 4-byte device unique code. The drift interference coefficient is stored in the 16-23 byte position (double-precision floating point format) in the data area, and the fault confidence level label is stored in the 24 byte position. The remaining byte areas are filled with default value zero. The report is encoded in little-endian byte sequence binary format, and a 4-byte CRC check code is added at the file header. The report generation frequency is set to 50 per second, and the report is output to the data analysis system through the gigabit Ethernet interface.

[0108] The abnormality processing mechanism covers three cases: when the detected projection scalar value exceeds the range [-1.50, +1.50], it is automatically limited to the interval [-1.00, +1.00] and then recalculated; when the fault feature vector two norm square is less than 0.10, the sample is discarded and the data re-sampling process is started; when the historical health running database is empty, the factory preset baseline value (default value is 0.50) is called. All abnormal events are recorded in detail logs, including timestamp, error code and processing result. The technical verification scheme is completed by injecting standard test vectors: the drift interference coefficient calculation error is ≤±1% (full scale), the fault confidence level determination accuracy is ≥99.95%, and the report output packet loss rate is <0.001%.

[0109] The technical scheme solves the problem of decoupling of cabinet signal drift and fault characteristics, and forms an organic whole through time-frequency double closed loop architecture, including interference environment simulation, double-track signal acquisition, drift quantization, characteristic injection, orthogonal decomposition, dynamic evaluation: in the interference environment simulation link, the pulse characteristics are dynamically bound with the physical response of switch operation; in the signal acquisition, the reference sensor and the monitoring unit are connected in parallel to realize true value reference; the feature band is extracted from the drift sequence as the fault excitation guide; the pollution degree of the drift to the fault is decoupled and quantized through the orthogonal projection of the feature vector; finally, the confidence evaluation system is adopted for dynamic calibration of historical reference. The technical chain of "interference simulation-drift quantization-fault injection-vector decoupling-dynamic evaluation" forms a systematic technical closed loop through multi-link collaborative innovation. Especially, the drift characteristic band is used as the fault injection reference to realize the precise matching of drift interference energy and fault excitation frequency.

[0110] Embodiment 2: Figure 2 The structure diagram of the comprehensive cabinet test device is given, and the comprehensive cabinet test device comprises:

[0111] The interference simulation module simulates a high-intensity transient electromagnetic interference environment in the cabinet.

[0112] The signal parallel module connects the reference sensor in parallel at the signal acquisition end of the signal monitoring unit, and synchronously acquires the output data of the reference sensor and the original analog signal of the signal monitoring unit.

[0113] The sequence generation module calculates the difference between the original analog signal of the signal monitoring unit and the output data of the reference sensor at the same time point, and generates a difference sequence.

[0114] The characteristic injection module extracts the feature band with concentrated interference energy in the difference sequence as the injection frequency reference, applies a preset fault excitation according to the injection frequency reference, and synchronously acquires the output data of the reference sensor and the original analog signal of the signal monitoring unit to generate a fault characteristic data set.

[0115] The vector decoupling module performs orthogonal decomposition based on the covariance matrix on the difference sequence and the fault characteristic data set, separates the drift characteristic vector and the fault characteristic vector, and calculates the projection scalar value of the drift characteristic vector in the direction of the fault characteristic vector.

[0116] The evaluation output module outputs the drift interference coefficient and the fault confidence report based on the drift characteristic vector, the fault characteristic vector and the projection scalar value.

[0117] The interference simulation module generates a transient electromagnetic environment conforming to the physical characteristics of switch operation, and provides a real interference background environment for the signal parallel module. The signal parallel module realizes millisecond-level synchronous acquisition of the output data of the reference sensor and the original analog signal of the signal monitoring unit under strong interference conditions through the electrical parallel topology, and outputs a double-track signal data stream. The sequence generation module performs time domain alignment processing on the double-track signal, and generates a difference sequence representing the drift characteristics through difference operation at the same timestamp point. The sequence is input into the subsequent module as a drift quantization reference. The innovation of the feature injection module lies in that after extracting the feature frequency band with concentrated interference energy from the difference sequence, the center frequency of the frequency band and the fault excitation parameter are dynamically bound, and a preset fault excitation with adaptive frequency is generated through a programmable signal generator; during the application of the excitation, the double-track signal is synchronously acquired to generate a fault feature data set, and closed-loop feedback of the drift feature to the fault excitation is realized. The vector decoupling module performs orthogonal decomposition of the covariance matrix on the difference sequence and the fault feature data set, and separates the fault feature vector (corresponding to the maximum eigenvalue) and the drift feature vector (corresponding to the second largest eigenvalue) by using the descending arrangement characteristics of the feature vector sequence; the projection scalar value of the drift feature vector in the direction of the fault feature vector is calculated through dot product operation, and the value quantifies the pollution degree of the drift to the fault signal. The evaluation output module generates a drift interference coefficient and a three-level confidence label based on the projection scalar value and the historical dynamic reference, outputs a diagnostic conclusion through a structured binary report, and realizes visual decision support of the drift-fault coupling relationship.

[0118] The calculations involved in the embodiments are all de-dimensioned numerical calculations, and the preset parameters and threshold values in the calculations are set by a person skilled in the art according to actual conditions.

[0119] It should be noted that the application can be deployed in the device itself to realize embedded application, or run on a PC or other terminal with a user interface, so as to meet various hardware environments and use requirements.

[0120] The above-described embodiments can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented by software, the above-described embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are wholly or partially generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions can be transferred from one website, computer, server, or data center to another website, computer, server, or data center through a wired (for example, infrared, wireless, microwave, etc.) manner. The computer-readable storage medium can be any available medium accessible by a computer or a data storage device such as a server, data center, etc. containing a set of one or more available media. The available media can be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state disk.

[0121] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described system, device and module can refer to the corresponding processes in the foregoing method embodiments, which will not be described here.

[0122] In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other ways. For example, the above-described device embodiments are only schematic, for example, the division of the modules is only a logical function division, and actual implementation can have another division manner, for example, a plurality of modules or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed each other can be indirect coupling or communication connection through some interfaces, devices or modules, which can be electrical, mechanical or other forms.

[0123] The modules described as separate components can or can not be physically separated, and the components displayed as modules can or can not be physical modules, which can be located in one place or distributed on a plurality of network modules. Part or all of the modules can be selected according to actual needs to achieve the purpose of the embodiments.

[0124] In addition, each functional module in the various embodiments of the present application can be integrated in one processing module, or each module can exist physically independently, or two or more modules can be integrated in one module.

[0125] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to make a computer device (which can be a personal computer, a server or a network device, etc.) execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0126] The above is merely specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0127] Finally: the above is only the preferred embodiments of the present application, and is not used to limit the present application, any modification, equivalent replacement, improvement, etc. within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A comprehensive cabinet testing method, characterized in that, include: S1. Simulate a high-intensity transient electromagnetic interference environment inside the cabinet; S2. Connect a reference sensor in parallel at the signal acquisition terminal of the signal monitoring unit to synchronously acquire the output data of the reference sensor and the original analog signal of the signal monitoring unit. S3. Calculate the difference between the original analog signal of the signal monitoring unit and the output data of the reference sensor at the same time point, and generate a difference sequence. S4. Extract the characteristic frequency band where the interference energy is concentrated in the difference sequence as the injection frequency reference, apply the preset fault excitation according to the injection frequency reference, and synchronously collect the output data of the reference sensor and the original analog signal of the signal monitoring unit to generate a fault feature dataset. S5. Perform orthogonal decomposition based on covariance matrix on the difference sequence and fault feature dataset to separate drift feature vector and fault feature vector, and calculate the projection scalar value of drift feature vector in the direction of fault feature vector. S6. Based on the drift feature vector, fault feature vector and projection scalar value, output the drift interference coefficient and fault confidence report.

2. The integrated cabinet testing method according to claim 1, characterized in that, Simulate a high-intensity transient electromagnetic interference environment within the server rack, including: A surge generator is used to inject current pulses into the power circuit of the switch control equipment in the cabinet. The rise time of the current pulse corresponds to the natural response time range of the high current switch action, the duration corresponds to the transient decay period caused by the switch operation, and the spectral characteristics of the current pulse match the typical switch operation spectrum defined in the standard interference waveform library.

3. The integrated cabinet testing method according to claim 1, characterized in that, A reference sensor is connected in parallel at the signal acquisition terminal of the signal monitoring unit to synchronously acquire the output data of the reference sensor and the original analog signal of the signal monitoring unit, including: The reference sensor signal input terminal is directly electrically connected in parallel with the signal acquisition circuit of the signal monitoring unit through conductive connectors. A dual-channel synchronous acquisition device is used to synchronously acquire the output data of the reference sensor and the original analog signal of the signal monitoring unit. One channel of the dual-channel acquisition device is connected to the output terminal of the reference sensor to acquire the output data of the reference sensor, and the other channel is connected to the analog signal terminal of the signal monitoring unit to acquire the raw analog signal of the signal monitoring unit. The transmission lines from the reference sensor output terminal to the acquisition device channel and from the analog signal terminal of the signal monitoring unit to the acquisition device channel are connected by shielded twisted-pair cables, and the difference in length between the two cables is controlled within a negligible range. The acquisition start command is synchronously initialized to both acquisition channels by an external trigger pulse.

4. The integrated cabinet testing method according to claim 1, characterized in that, Calculate the difference between the original analog signal from the signal monitoring unit and the output data from the reference sensor at the same time point, and generate a difference sequence, including: Acquire the data sequence of the reference sensor output data and the original analog signal from the signal monitoring unit stored in the dual-channel synchronous acquisition device; The time-domain alignment processing of the data sequences of the reference sensor output data and the original analog signal of the signal monitoring unit adopts the same timestamp matching mechanism; At the timestamp when the matching is completed, the difference operation is performed, and the operation objects are the instantaneous value of the original analog signal of the signal monitoring unit and the instantaneous value of the output data of the reference sensor; The difference calculation results are stored sequentially with timestamps as independent variables to generate a difference sequence; Before the difference sequence is transmitted to the buffer in real time, zero-mean preprocessing is performed to suppress DC component drift interference.

5. The integrated cabinet testing method according to claim 1, characterized in that, The characteristic frequency bands where interference energy is concentrated in the difference sequence are extracted as the injection frequency reference, including: Perform a Fourier transform on the difference sequence to obtain spectral distribution data, and identify frequency bands with spectral amplitudes exceeding ten times the average energy level as characteristic frequency bands where interference energy is concentrated. The center frequency of the characteristic frequency band is set as the injection frequency reference, and a preset fault excitation signal centered on the corresponding injection frequency reference is generated by a programmable signal generator.

6. The integrated cabinet testing method according to claim 1, characterized in that, A preset fault excitation is applied according to the injection frequency reference, and the output data of the reference sensor and the original analog signal of the signal monitoring unit are collected simultaneously to generate a fault feature dataset, including: The preset fault excitation signal is connected to the signal injection port of the switch control equipment in the cabinet via a power amplifier; During the application of the preset fault excitation, the acquisition operations of the reference sensor output data and the original analog signal of the signal monitoring unit are performed synchronously, and the acquisition time window covers the entire excitation cycle; The collected benchmark sensor output data and the original analog signal from the signal monitoring unit are timestamped and encapsulated to generate a fault feature dataset.

7. The integrated cabinet testing method according to claim 1, characterized in that, An orthogonal decomposition based on the covariance matrix is ​​performed on the difference sequence and the fault feature dataset to separate the drift feature vector and the fault feature vector. The projection scalar value of the drift feature vector onto the direction of the fault feature vector is calculated, including: Construct a joint sample matrix of difference sequence data points and fault feature dataset data points, and align them according to the same timestamp to form a two-dimensional data array; Calculate the covariance matrix of the two-dimensional data array. The element values ​​of the covariance matrix represent the degree of statistical correlation between the difference sequence data and the fault characteristic data. Performing eigenvalue decomposition on the covariance matrix yields a sequence of eigenvectors arranged in descending order of their eigenvalues. The eigenvector corresponding to the largest eigenvalue is defined as the fault eigenvector; the eigenvector corresponding to the second largest eigenvalue is defined as the drift eigenvector. The projection scalar value is calculated by the dot product operation of the drift feature vector and the fault feature vector. The result of the dot product operation represents the projection length of the drift feature vector in the direction of the fault feature vector.

8. The integrated cabinet testing method according to claim 1, characterized in that, Based on the drift feature vector, fault feature vector, and projection scalar value, the system outputs a drift interference coefficient and fault confidence report, including: The drift interference coefficient is defined by processing the result of the exponential function of the projected scalar value. The base of the exponential function is set to a natural constant and the exponent term is a negative proportional amplification of the projected scalar value. The baseline value of fault confidence is calculated based on the square of the second norm of the fault feature vector, where the square of the second norm represents the energy accumulation level of the fault feature vector. The relative confidence ratio is obtained by comparing the fault confidence baseline value with the preset historical healthy operation period baseline value; The fault confidence level label is marked according to the interval mapping relationship between the relative confidence ratio and the preset threshold; The drift interference coefficient is combined with the fault confidence level label to generate a structured drift interference coefficient and fault confidence report.

9. A comprehensive cabinet testing device, used to implement the comprehensive cabinet testing method according to any one of claims 1-8, characterized in that, include: Interference simulation module: Simulates a high-intensity transient electromagnetic interference environment within the cabinet; Signal parallel module: A reference sensor is connected in parallel to the signal acquisition terminal of the signal monitoring unit to synchronously acquire the output data of the reference sensor and the original analog signal of the signal monitoring unit; Sequence generation module: Calculates the difference between the original analog signal of the signal monitoring unit and the output data of the reference sensor at the same time point, and generates a difference sequence; Feature injection module: Extracts the characteristic frequency bands of interference energy concentration in the difference sequence as the injection frequency reference, applies a preset fault excitation according to the injection frequency reference, and synchronously collects the output data of the reference sensor and the original analog signal of the signal monitoring unit to generate a fault feature dataset; Vector decoupling module: Performs orthogonal decomposition based on covariance matrix on difference sequence and fault feature dataset, separates drift feature vector and fault feature vector, and calculates the projection scalar value of drift feature vector in the direction of fault feature vector; Evaluation output module: Based on the drift feature vector, fault feature vector and projected scalar value, output drift interference coefficient and fault confidence report.

Images