Methods and systems for processing operational status data of electronic circuit boards

Abstract

This invention provides a method and system for processing operational status data of electronic circuit boards. By acquiring real-time voltage fluctuation data of each power path of the electronic circuit board within a preset monitoring period, adaptive time-frequency decomposition is performed to separate the steady-state basis component characterizing the stable operation characteristics of the circuit board and the instantaneous pulsation component reflecting transient interference phenomena. Based on the amplitude change sequence and the time interval parameter of the corresponding sampling timestamp, an amplitude-time series correlation heatmap is constructed in a two-dimensional data plane. This heatmap is then fused pixel-by-pixel with the standard pulsation pattern diagram stored at the time of the circuit board's manufacture to generate a fused difference spectrum containing the spatial distribution characteristics of the differing pixels. The state coupling coefficient of each power path is calculated based on the spatial density distribution characteristics of the differing pixels. The state coupling coefficients of all power paths are then integrated to obtain a quantified condensation index. This invention can improve the reliability and comprehensiveness of operational status assessment.

Description

Technical Field

[0001] This application relates to the field of data processing, and in particular to a method and system for processing operating status data of electronic circuit boards. Background Technology

[0002] In the field of electronic equipment manufacturing and maintenance, monitoring the operational status of electronic circuit boards is a crucial foundation for ensuring reliable equipment operation. Currently, the industry generally relies on voltage amplitude comparisons based on preset thresholds or simple statistical characteristic calculations. This involves comparing real-time collected voltage values ​​with fixed thresholds or calculating statistical quantities such as the mean and variance of voltage data to determine if there are any anomalies on the circuit board. However, these methods are often limited to isolated analysis of voltage data at a single point in time or within a local time period. They struggle to effectively distinguish between inherent fluctuations within the normal operating range of the circuit board and abnormal interference caused by component aging, poor contact, or other issues. Furthermore, they fail to capture the state correlations between different power paths, leading to a one-sided assessment of the overall operational health of the circuit board and a tendency for misjudgments or omissions. Summary of the Invention

[0003] This invention provides a method and system for processing operational status data of electronic circuit boards.

[0004] In a first aspect, embodiments of the present invention provide a method for processing operational status data of an electronic circuit board. The method includes: acquiring real-time voltage fluctuation data of each power path of the electronic circuit board within a preset monitoring period; the real-time voltage fluctuation data is synchronously acquired at a preset sampling frequency by high-frequency sampling units distributed at key power nodes of the circuit board, and includes a continuous time series of voltage amplitude change records; performing adaptive time-frequency decomposition on the real-time voltage fluctuation data to separate a steady-state base component characterizing the stable operation characteristics of the circuit board and an instantaneous pulsation component reflecting transient interference phenomena; the steady-state base component is a trend fitting curve of the voltage fluctuation data on the time axis, and the instantaneous pulsation component is a difference sequence between the original data and the steady-state base component; and based on the amplitude change sequence of the instantaneous pulsation component and the corresponding sampling... The time interval parameter of the timestamp is used to construct an amplitude-time series correlation heatmap in a two-dimensional data plane. The horizontal axis of the heatmap corresponds to the time distribution of the sampling timestamp, and the vertical axis corresponds to the amplitude interval division of the instantaneous pulsation component. The gray value of the pixel in the graph is positively correlated with the frequency of pulsation at the corresponding time-amplitude coordinate. The amplitude-time series correlation heatmap is morphologically fused pixel by pixel with the standard pulsation pattern map stored at the time of the circuit board leaving the factory. By calculating the difference in gray value of the corresponding pixel, a fused difference spectrum containing the spatial distribution characteristics of the difference pixels is generated. The state coupling coefficient of each power path is calculated according to the spatial density distribution characteristics of the difference pixels in the fused difference spectrum. The state coupling coefficients of all power paths are integrated according to a preset weighting rule to obtain a quantitative condensed index characterizing the overall operating health status of the circuit board.

[0005] Secondly, embodiments of the present invention provide a computer system, including: a memory for storing computer-executable instructions or computer programs; and a processor for executing the computer-executable instructions or computer programs stored in the memory to implement the above-described method for processing operating status data applied to electronic circuit boards.

[0006] The embodiments of this application have the following beneficial effects:

[0007] This invention acquires real-time voltage fluctuation data of each power path within a preset monitoring period and performs adaptive time-frequency decomposition to separate steady-state base components and instantaneous pulsation components. This accurately separates trend components reflecting the stable operation characteristics of the circuit board from dynamic fluctuation components exhibiting instantaneous interference, avoiding the limitations of conventional fixed-frequency filtering in distinguishing between long-term trends and instantaneous interference. Based on the amplitude change sequence of the instantaneous pulsation component and the time interval parameter of the sampling timestamp, an amplitude-time series correlation heatmap is constructed, transforming one-dimensional voltage fluctuation data into a two-dimensional spatiotemporal representation. This intuitively presents the clustering patterns of instantaneous pulsations in time distribution and amplitude range, breaking through the limitations of traditional numerical statistics which can only reflect... This approach overcomes the limitations of single-point fluctuations. By fusing the amplitude-time correlation heatmap with the standard pulsation pattern map pixel by pixel to generate a fusion difference spectrum, it can comprehensively capture the subtle spatial differences between the real-time pulsation pattern and the factory standard pattern, reflecting the regional distribution characteristics of abnormal pulsations. Based on the spatial density distribution characteristics of the difference pixels in the fusion difference spectrum, the state coupling coefficient of each power path is calculated and integrated into a quantitative condensation index, enabling a comprehensive consideration of the state correlation between multiple power paths. This avoids the inadequacy of single-path evaluation in reflecting the overall operating status, thereby achieving an accurate quantitative assessment of the overall operating health status of the electronic circuit board and improving the reliability and comprehensiveness of the operating status assessment. Attached Figure Description

[0008] Figure 1 This is a schematic diagram of the structure of the computer system provided in the embodiments of this application;

[0009] Figure 2 This is a flowchart illustrating the method for processing operational status data of electronic circuit boards provided in this application embodiment. Detailed Implementation

[0010] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0011] The following describes a computer system for implementing the operational status data processing method for electronic circuit boards provided in the embodiments of this application. See also Figure 1This is a schematic diagram of the structure of a computer system provided in an embodiment of this application. The computer system includes at least one processor 210, a memory 250, at least one network interface 220, and an external interface 230. The various components in the computer system 200 are coupled together through a bus system 240. It is understood that the bus system 240 is used to implement communication between these components. In addition to a data bus, the bus system 240 also includes a power bus, a control bus, and a status signal bus. However, for clarity, ... Figure 1 The general labeled all buses as Bus System 240.

[0012] Processor 210 can be an integrated circuit chip with signal processing capabilities, such as a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor can be a microprocessor or any conventional processor, etc.

[0013] External interface 230 may include, for example, one or more speakers and / or one or more visual displays. External interface 230 may also include one or more input devices 432, such as a keyboard, mouse, microphone, touch screen display, camera, etc.

[0014] The memory 250 may be removable, non-removable, or a combination thereof. Exemplary hardware devices include solid-state storage, hard disk drives, optical disk drives, etc. The memory 250 may optionally include one or more storage devices physically located away from the processor 210.

[0015] The memory 250 may include volatile memory or non-volatile memory, or both. The non-volatile memory may be read-only memory (ROM), and the volatile memory may be random access memory (RAM). The memory 250 described in this application embodiment is intended to include any suitable type of memory.

[0016] In some embodiments, memory 250 is capable of storing data to support various operations, examples of which include programs, modules, and data structures or subsets or supersets thereof, as illustrated below.

[0017] Operating system 251 includes system programs for handling various basic system services and performing hardware-related tasks, such as the framework layer, core library layer, driver layer, etc., for implementing various basic business functions and handling hardware-based tasks;

[0018] The network communication module 252 is used to reach a data acquisition device (such as a high-frequency sampling unit) via one or more (wired or wireless) network interfaces 220. Exemplary network interfaces 220 include Bluetooth, WiFi, and Universal Serial Bus (USB), etc.

[0019] Presentation module 253 is configured to enable the display of information (e.g., external interface for operating peripheral devices and displaying content and information) via one or more output devices 231 (e.g., display screen, speaker, etc.) associated with external interface 230;

[0020] The input processing module 254 is used to detect and translate one or more user inputs or interactions from one or more input devices 232.

[0021] The following describes the method for processing operational status data of electronic circuit boards provided in the embodiments of this application. In actual implementation, the method for processing operational status data of electronic circuit boards provided in the embodiments of this application can be implemented by a computer system. See also Figure 2 , Figure 2 This is a flowchart illustrating the method for processing operational status data of electronic circuit boards provided in this application embodiment. Next, it will be discussed in conjunction with... Figure 2 The steps shown are explained.

[0022] Step S100: Obtain real-time voltage fluctuation data of each power path of the electronic circuit board within a preset monitoring period. The real-time voltage fluctuation data is synchronously obtained by high-frequency sampling units distributed on key power nodes of the circuit board at a preset sampling frequency, and includes a record of voltage amplitude changes in a continuous time series.

[0023] Real-time voltage fluctuation data reflects the dynamic changes in voltage across various power paths of an electronic circuit board over time during operation, presented as a continuous time-series record of voltage amplitude changes. The preset monitoring duration is a pre-defined time range for monitoring circuit board voltage. The high-frequency sampling unit is installed on critical power nodes of the circuit board. These critical power nodes typically play a crucial role in the normal operation of the circuit board, and voltage changes significantly impact overall performance, such as power supply pins of major chips and power input / output interfaces. The preset sampling frequency determines the number of times voltage data is collected per unit time; a higher sampling frequency allows for more accurate capture of rapid voltage changes. The high-frequency sampling unit uses high-precision voltage sensors, installed at various critical power nodes on the circuit board. Within the preset monitoring duration, the high-frequency sampling unit operates continuously, acquiring voltage data at the preset sampling frequency, ultimately obtaining a continuous time-series record of voltage amplitude changes. This data is stored in the data storage module.

[0024] Step S200: Perform adaptive time-frequency decomposition on the real-time voltage fluctuation data to separate the steady-state base component, which characterizes the stable operation of the circuit board, and the instantaneous pulsation component, which reflects the instantaneous interference phenomenon. The steady-state base component is the trend fitting curve of the voltage fluctuation data on the time axis, and the instantaneous pulsation component is the difference sequence between the original data and the steady-state base component.

[0025] In one implementation, step S200 may include the following steps S210 to S260:

[0026] Step S210: Perform time series characteristic scanning on real-time voltage fluctuation data, extract different time windows according to preset sliding step size, extract the fluctuation period characteristics and amplitude change rate of voltage data in each window, and generate a fluctuation characteristic spectrum containing period-amplitude correlation. The sliding step size of the window is the same as the reciprocal of the sampling frequency.

[0027] Time series characteristic scanning involves a detailed analysis of real-time voltage fluctuation data in chronological order to uncover hidden periodicity and patterns of change. A preset sliding step size determines the interval at which the time window moves along the time axis; its size is the reciprocal of the sampling frequency. This setting ensures that the window movement matches the data sampling frequency, enabling accurate segmented analysis of the data. A time window is a range of data intercepted on the time axis. By analyzing data within different time windows, the specific characteristics of the voltage data in each time period can be obtained. Fluctuation period characteristics describe the changes in voltage data within a complete cycle, including the length of the cycle and the shape of the waveform; the amplitude change rate represents the proportion of voltage amplitude change per unit time, reflecting the speed of voltage change.

[0028] Step S220: Project statistics on the frequency axis of the wave characteristic spectrum, mark the frequency point where the cumulative value of the amplitude change rate forms the peak, expand outward from the peak point to both sides until the cumulative value drops to the median level of the cumulative value corresponding to the peak, and mark the expanded interval as the inherent interference frequency interval.

[0029] Projective statistics is an analytical method for analyzing fluctuation characteristic spectra on the frequency axis. By statistically analyzing the cumulative amplitude change rate at each frequency point, it identifies frequencies where interference may occur. The cumulative amplitude change rate is the value obtained by summing the amplitude change rates at each frequency point within a certain range on the frequency axis, reflecting the overall amplitude variation near that frequency point. Peak points are the frequency points where the cumulative amplitude change rate reaches its maximum value. These peak points are usually related to potential interference frequencies during circuit board operation because voltage amplitude changes are often more drastic at interference frequencies, leading to peak values ​​in the cumulative values. The inherent interference frequency range is a frequency range defined around the peak points. Within this range, voltage fluctuations may be caused by inherent interference factors within the circuit board itself.

[0030] In one implementation, step S220 may specifically include the following steps S221 to S226:

[0031] Step S221: Scan the frequency axis of the wave characteristic spectrum point by point, count the cumulative value of the amplitude change rate corresponding to each frequency point, and generate a frequency-cumulative value distribution curve. The peak point of the curve corresponds to the potential interference frequency center.

[0032] Point-by-point scanning involves analyzing the frequency axis of the fluctuation characteristic spectrum sequentially to obtain detailed information for each frequency point. The cumulative amplitude change rate is the result of accumulating the amplitude change rate corresponding to each frequency point within a certain frequency range, comprehensively reflecting the overall trend of amplitude change near that frequency point. The frequency-cumulative value distribution curve uses frequency as the horizontal axis and the cumulative amplitude change rate as the vertical axis. By plotting this curve, the distribution of cumulative values ​​corresponding to each frequency point can be visually displayed. The potential interference frequency center refers to the frequency point on the frequency-cumulative value distribution curve where the cumulative value reaches its peak. These frequency points are often the main frequency locations where interference may occur during circuit board operation, because at the interference frequency, the voltage amplitude change is usually more significant, leading to a peak in the cumulative value.

[0033] Step S222: Traverse the frequency-cumulative value distribution curve. When the cumulative value of multiple consecutive frequency points shows a trend of first rising and then falling, mark the trend inflection point as the peak point and record the frequency value and corresponding cumulative value of the peak point.

[0034] Traversing the frequency-cumulative value distribution curve means conducting a comprehensive examination and analysis of each data point on the curve. A cumulative value that first rises and then falls across multiple consecutive frequency points indicates that the cumulative amplitude change rate exhibits a trend of first increasing and then decreasing within this frequency range. The turning point of this trend reflects that the cumulative value has reached a local maximum. The peak point is this turning point, representing a relatively high level of cumulative amplitude change rate near that frequency location, usually associated with potential interference frequencies. Recording the frequency value and corresponding cumulative value of the peak point is crucial for accurately determining the specific location and intensity of the interference frequency, enabling in-depth analysis and processing of interference on the circuit board.

[0035] In practice, analysts start from the beginning of the frequency-cumulative value distribution curve and examine the cumulative value of each data point sequentially in ascending order of frequency. When they find that the cumulative values ​​of multiple consecutive data points show a trend of first rising and then falling, they identify the data point where this trend reverses and mark it as the peak point. Then, they accurately record the frequency and cumulative value corresponding to that peak point.

[0036] Step S223: Starting from the frequency value of each peak point, scan the frequency-cumulative value distribution curve point by point in the low-frequency direction until the cumulative value drops to the median level of the peak cumulative value, and record the frequency value at that position as the low-frequency boundary; perform the same scan in the high-frequency direction and record the high-frequency boundary.

[0037] Scanning from the peak frequency is to determine the boundaries of the inherent interference frequency range centered on that peak. The low-frequency direction is the direction of gradually decreasing frequency, and the high-frequency direction is the direction of gradually increasing frequency. Point-by-point scanning involves sequentially checking the cumulative value of each frequency point on the frequency-cumulative value distribution curve. The median level of the peak cumulative value is half the cumulative value corresponding to the peak point, serving as an important reference value for determining the boundaries. The low-frequency and high-frequency boundaries define the possible ranges of interference frequencies in the low-frequency and high-frequency directions, respectively.

[0038] For example, for each peak point marked in step S222, its frequency value is first located. Then, starting from this frequency value, the frequency-cumulative value distribution curve is scanned point by point towards the lower frequency direction, continuously comparing the cumulative value of the current frequency point with the median level of the peak cumulative value. When the cumulative value drops to the median level of the peak cumulative value, the frequency value at that location is recorded; this frequency value is the low-frequency boundary. Next, starting from the frequency value of this peak point, the same point-by-point scanning operation is performed towards the higher frequency direction. When the cumulative value drops to the median level of the peak cumulative value again, the frequency value at that location is recorded; this is the high-frequency boundary.

[0039] Step S224: Mark the frequency range between the low-frequency boundary and the high-frequency boundary as the candidate interference frequency interval. Each peak point corresponds to a candidate interval, and the range of the candidate interval is defined by the low-frequency boundary and the high-frequency boundary.

[0040] For example, after obtaining the low-frequency boundary and high-frequency boundary corresponding to each peak point through step S223, the frequency range between these two boundaries is marked as a candidate interference frequency interval. For instance, for a peak point, its low-frequency boundary is a relatively low frequency value, and its high-frequency boundary is a relatively high frequency value; then the frequency range between these two frequency values ​​is marked as a candidate interference frequency interval. By performing this processing on all peak points, a series of candidate interference frequency intervals are obtained. These intervals form the basis for further analysis of circuit board interference. Subsequently, these candidate intervals can be screened and evaluated to determine which intervals are the inherent interference frequency intervals that truly affect the operation of the circuit board.

[0041] Step S225: Read the normal interference bandwidth range in the circuit board design document, compare the bandwidth of the candidate interference frequency interval with the normal range, and mark the candidate interval with bandwidth exceeding the normal range as the inherent interference frequency interval that needs to be focused on.

[0042] For example, the operator reads the normal interference bandwidth range from the circuit board design document. For each candidate interference frequency interval marked in step S224, its bandwidth, i.e., the difference between the high-frequency boundary and the low-frequency boundary, is calculated. Then, this bandwidth is compared with the normal interference bandwidth range. If the bandwidth of a candidate interference frequency interval exceeds the normal range, the interval is marked as an inherent interference frequency interval requiring special attention.

[0043] Step S226: Perform an overlap check on the marked inherent interference frequency intervals. If the boundaries of adjacent intervals overlap, merge the overlapping parts into a single interval. The boundaries of the merged interval are the minimum low-frequency boundary and the maximum high-frequency boundary of the original interval, generating the final set of inherent interference frequency intervals.

[0044] Overlap checking involves a comprehensive examination of all marked inherent interference frequency ranges requiring close monitoring, checking for any overlap between adjacent ranges. Adjacent ranges refer to two inherent interference frequency ranges that are geographically adjacent on the frequency axis. If the boundaries of adjacent ranges overlap, it indicates that the two ranges cover partially the same frequency range, potentially representing the same or related interference source. Merging them into a single range provides a more accurate representation of the interference frequency range. The boundaries of the merged range are taken from the minimum low-frequency boundary and the maximum high-frequency boundary of the original range. This ensures that the merged range completely covers the overlapping portions of the original two ranges, as well as their independent portions, avoiding any omissions. The final set of inherent interference frequency ranges is the collection of all inherent interference frequency ranges obtained after overlap checking and merging. This set clearly reflects the range of interference frequencies that may exist during circuit board operation.

[0045] Step S230: Based on the number of inherent interference frequency intervals and the total bandwidth, determine the number of frequency bands for time-frequency decomposition, so that each interval corresponds to an independent frequency band, and the ratio of the number of frequency bands to the number of intervals is not less than a preset ratio. At the same time, insert transition frequency bands between adjacent intervals, and the bandwidth of the transition frequency band is the average value of the bandwidth of the adjacent intervals.

[0046] The number of frequency bands in time-frequency decomposition refers to the number of frequency ranges divided during adaptive time-frequency decomposition. Determining the appropriate number of frequency bands is crucial for accurately analyzing voltage fluctuation data. The number of inherent interference frequency intervals reflects the number of different frequency regions where interference may exist, while the total bandwidth reflects the overall width of all interference frequency ranges. Having each interval correspond to an independent frequency band means that each inherent interference frequency interval can be individually represented in the frequency band division, allowing for more detailed analysis of the voltage data characteristics within each interference frequency interval. The ratio of the number of frequency bands to the number of intervals is not less than a preset ratio to ensure that the frequency band division is sufficiently detailed and can fully cover all inherent interference frequency intervals. The preset ratio is determined based on experience and actual analysis needs. Transition frequency bands are inserted between adjacent inherent interference frequency intervals, with a bandwidth equal to the average bandwidth of the adjacent intervals. The role of transition frequency bands is to smoothly transition between adjacent frequency bands, reducing interference and data jumps between bands, making the time-frequency decomposition results more accurate and stable.

[0047] Step S240: Input the real-time voltage fluctuation data into the time-frequency decomposition process according to the determined frequency band parameters, divide the data into multi-resolution frequency bands, and generate energy distribution matrices for different frequency bands. The row dimension of the matrix corresponds to the time axis, the column dimension corresponds to the frequency axis, and the element value represents the energy density of the corresponding time-frequency point.

[0048] Real-time voltage fluctuation data contains information on the voltage changes of each power path on the electronic circuit board over time during operation. The determined frequency band parameters are the number of frequency bands, band boundaries, and transition bands determined in step S230 based on the inherent interference frequency range. The time-frequency decomposition process is a procedure that converts time-domain signals to the frequency domain for analysis, allowing real-time voltage fluctuation data to be separated according to different frequency components. Multi-resolution frequency band division divides the frequency range into multiple frequency bands of different resolutions during the time-frequency decomposition process, enabling more detailed analysis of voltage data characteristics in different frequency ranges. The energy distribution matrix is ​​a two-dimensional matrix with rows corresponding to the time axis and columns corresponding to the frequency axis. Each element in the matrix represents the energy density at the corresponding time-frequency point, reflecting the amount of energy contained in the voltage signal at that time and frequency.

[0049] In one implementation, step S240 may specifically include the following steps S241 to S246:

[0050] Step S241: Based on the bandwidth of each interval in the inherent interference frequency interval set and the historical fault association records, calculate the dynamic influence factor of each interval. The larger the bandwidth and the more frequently the interval appears in the historical fault association records, the larger the influence factor. Perform time-domain segmentation enhancement processing on the real-time voltage fluctuation data through the interval influence factor.

[0051] For example, firstly, for each interval in the set of inherent interference frequency intervals, its bandwidth and frequency of occurrence in historical fault association records are statistically analyzed. For instance, historical fault association records can be obtained through database queries, and the frequency of occurrence for each interval can be counted. Then, according to a preset calculation rule, the dynamic impact factor for each interval is calculated by combining bandwidth and frequency of occurrence. The calculation rule can be a weighted summation formula, where bandwidth and frequency of occurrence are assigned different weights, the magnitude of which is determined based on practical experience and understanding of the circuit board. After obtaining the dynamic impact factor for each interval, the real-time voltage fluctuation data is segmented in the time domain. The data is divided into multiple time periods in chronological order, each time period corresponding to a different inherent interference frequency interval. Then, the voltage data within that time period is enhanced based on the dynamic impact factor of each interval. For example, the voltage data can be amplified or subjected to other weighted processing methods to make the voltage data corresponding to intervals with larger dynamic impact factors more prominent in the overall data.

[0052] Step S242: Based on the number of inherent interference frequency intervals and frequency distribution characteristics, construct a multi-resolution decomposition frequency band tree structure. The hierarchical depth of the tree structure is positively correlated with the number of intervals. Each level corresponds to a frequency band. The frequency band boundaries between levels are determined by the frequency axis bisection method of the previous level, so that the frequency band division covers all inherent interference frequency intervals.

[0053] For example, a frequency band tree structure is constructed based on the number and frequency distribution characteristics of inherent interference frequency ranges. First, the entire frequency range is used as the root node of the tree structure. Then, the initial hierarchical depth of the tree structure is determined based on the number of ranges. As the number of levels increases, the frequency bands of each higher level are divided using a frequency axis bisection method. For example, for a higher level frequency band, its frequency range is divided into two equal parts on the frequency axis, serving as two sub-bands of the next level. During the division process, it is continuously checked whether the divided frequency bands cover all inherent interference frequency ranges; if any uncovered ranges exist, the division method is further adjusted. The tree structure is constructed layer by layer in this way until the preset hierarchical depth is reached.

[0054] Step S243: For each level of the tree structure, the filter parameters of that level are dynamically configured according to the start frequency, end frequency and isolation requirements of the corresponding frequency band. The passband range of the filter matches the frequency boundary of the frequency band, and the stopband range avoids the frequency range of the adjacent frequency band, so as to perform frequency band isolation processing on the real-time voltage fluctuation data.

[0055] Each level of the tree structure corresponds to a frequency band, and each band has its own start and end frequencies. These frequency values ​​determine the range of that band on the frequency axis. The isolation requirement for adjacent frequency bands means that to avoid interference between signals from different bands, appropriate parameters need to be set in the filter design to effectively separate signals from different bands. Dynamically configuring filter parameters involves adjusting the relevant filter parameters in real time based on the frequency band characteristics and isolation requirements of each level to achieve optimal frequency band isolation. The passband range of a filter refers to the frequency range that signals are allowed to pass through. This range matches the frequency boundaries of the corresponding band, ensuring that signals in that band can pass through the filter smoothly. The stopband range of a filter refers to the frequency range that signals are blocked from passing through. This range must avoid the frequency ranges of adjacent bands to reduce crosstalk between adjacent bands. Frequency band isolation processing uses configured filters to filter real-time voltage fluctuation data, separating signals from different bands for subsequent individual analysis of each band's signal.

[0056] Step S244: For the voltage data that has undergone frequency band isolation processing, the signal components are extracted sequentially according to the hierarchical order. Each extracted signal component corresponds to an inherent interference frequency range. During the extraction process, the crosstalk between the current frequency band and the frequency bands above and below is monitored in real time. The crosstalk is calculated by the ratio of the signal power leakage to the signal power of the current frequency band.

[0057] For example, starting from the root level of the frequency band tree structure, signal components are extracted from the frequency band-isolated voltage data at each level. Appropriate signal processing methods, such as Fourier transform and wavelet transform, can be used to convert the voltage data from the time domain to the frequency domain and extract the signal components. During the extraction process, the signal crosstalk between the current frequency band and the frequency bands above and below is monitored in real time. First, the signal power of the current frequency band is calculated, which can be obtained by squaring and integrating the voltage signal within the current band. Then, the signal power leaked from adjacent frequency bands into the current frequency band is calculated, for example, by analyzing the energy distribution of signals from adjacent frequency bands within the current frequency band to estimate the signal power leakage. Finally, the signal power leakage is divided by the signal power of the current frequency band to obtain the signal crosstalk. For example, in the signal component extraction process of a certain level, the calculated signal power of the current frequency band is a fixed value, and the signal power leaked from adjacent frequency bands into the current frequency band is another value; dividing the two yields the signal crosstalk.

[0058] Step S245: When the crosstalk exceeds the preset threshold, adjust the transition band slope of the current level filter. The transition band slope is positively correlated with the crosstalk. At the same time, backtrack and optimize the frequency band boundary of the previous level, expand or shrink the isolation band width of adjacent frequency bands until the crosstalk drops below the threshold.

[0059] Crosstalk exceeding a preset threshold means that the signal interference between the current frequency band and adjacent frequency bands is too great, exceeding the acceptable range, and adjustment measures are needed. The preset threshold is a critical value set based on the circuit board design requirements and actual operating experience. When crosstalk exceeds this value, it affects the accurate analysis and judgment of the circuit board's voltage signals. Adjusting the transition band slope of the current level filter is one method to reduce signal crosstalk. The transition band slope refers to the slope of the filter in the transition region between the passband and stopband. The larger the slope, the faster the filter attenuates in the transition region, and the better it can prevent signals from adjacent frequency bands from entering the current frequency band. The transition band slope is positively correlated with crosstalk; that is, the greater the crosstalk, the larger the transition band slope should be adjusted. Backtracking and optimizing the frequency band boundary of the previous level means that when the crosstalk problem of the current level cannot be completely solved by adjusting the filter's transition band slope, returning to the previous level and readjusting the frequency band boundary, expanding or narrowing the isolation band width of adjacent frequency bands, to further reduce crosstalk between frequency bands. By continuously adjusting until the crosstalk drops below the threshold, the accuracy and reliability of signal extraction are ensured.

[0060] Step S246: For the signal components extracted at each level, calculate the square of the signal amplitude at each sampling point in the order of the time axis as the energy density value. Arrange the energy density values ​​of different levels in the order of frequency bands to obtain a two-dimensional matrix with time as the row and frequency band as the column. The matrix element values ​​reflect the degree of energy concentration at the corresponding time-frequency band points.

[0061] For example, for each level of extracted signal components, each sampling point is processed sequentially according to the time axis. For each sampling point, the square of its signal amplitude is calculated to obtain the energy density value of that sampling point. For example, for a sampling point at a certain level, its signal amplitude is a specific value; this value is squared to obtain the energy density value of that sampling point. Then, the energy density values ​​of different levels are arranged in frequency band order. The order of the frequency bands in the frequency band tree structure can be determined first, and then the energy density values ​​of each frequency band at different time points are sequentially filled into the corresponding positions in the matrix. For example, for a time point and a frequency band, the energy density value of that frequency band at that time point is filled into the intersection of the corresponding row and column in the matrix. In this way, a two-dimensional matrix is ​​constructed with time as the row and frequency band as the column. By observing the magnitude and distribution of the element values ​​in this matrix, the energy accumulation areas of the circuit board voltage signal at different times and frequency bands can be analyzed, thereby determining whether there are abnormal energy fluctuations on the circuit board, as well as the frequency range and time points that these fluctuations may correspond to, which helps to further investigate circuit board faults and interference problems.

[0062] Step S250: Define a preset low-frequency band range on the frequency axis of the energy distribution matrix. Within this range, identify regions where the energy density is higher than the average level of adjacent regions and the fluctuation amplitude is less than a preset fluctuation threshold within multiple consecutive time windows. Mark these regions as the base signal frequency band. Perform an inverse transformation on the energy distribution of the base signal frequency band to obtain the preliminary extraction results of the steady-state base component. Define a preset high-frequency band range on the frequency axis. Identify regions where the rate of change of energy density within a single time window exceeds the average rate of change of adjacent time windows. Mark these regions as the interference signal frequency band. Perform noise reduction processing on the energy distribution of the interference signal frequency band to obtain the preliminary extraction results of the instantaneous pulsation component.

[0063] The preset high-frequency band range is a pre-defined high-frequency region on the frequency axis. Signal energy changes in this high-frequency region are typically drastic, potentially containing interference signals from the circuit board's operation. Within this range, regions where the rate of energy density change exceeds the average rate of change in adjacent time windows are identified. These regions represent rapid changes in signal energy within that time window, possibly due to transient interference, and are marked as interference signal frequency bands. Noise reduction processing modifies the energy distribution of the interference signal frequency band to remove noise components, allowing for more accurate extraction of transient pulsation components. By performing noise reduction processing on the energy distribution of the interference signal frequency band, preliminary extraction results of the transient pulsation components are obtained, reflecting the transient interference occurring during circuit board operation.

[0064] For example, firstly, a preset low-frequency band and a preset high-frequency band are defined on the frequency axis of the energy distribution matrix. Then, for the preset low-frequency band, the time axis is divided into multiple time windows, the energy density within each time window is calculated, and compared with the average energy density of adjacent regions, while checking whether the fluctuation amplitude is less than a preset fluctuation threshold. If multiple consecutive time windows meet the condition, the region is marked as the base signal frequency band. The energy distribution of this frequency band is inversely transformed, for example, using inverse Fourier transform, to convert it back to a time-domain signal, obtaining the preliminary extraction result of the steady-state base component. For the preset high-frequency band, the rate of change of energy density in each individual time window is calculated and compared with the average rate of change of adjacent time windows. If the rate of change of energy density in a certain time window exceeds the average rate of change of adjacent time windows, the region is marked as the interference signal frequency band. The energy distribution of this frequency band is denoised, for example, using filtering, smoothing, or other methods to remove noise, obtaining the preliminary extraction result of the instantaneous pulsation component.

[0065] Step S260: Superimpose the preliminary results of the steady-state base component and the preliminary results of the instantaneous pulsation component in the time domain, compare the superimposed results with the original real-time voltage fluctuation data, calculate the mean square error of the two, and if the error exceeds the preset range, readjust the frequency band parameters and repeat the decomposition process until the error meets the requirements, and generate the final steady-state base component and instantaneous pulsation component.

[0066] For example, the preliminary results of the steady-state base component and the preliminary results of the instantaneous pulsation component are added in the time domain to obtain the superimposed signal. Then, the difference between the superimposed signal and the original real-time voltage fluctuation data is calculated point by point at each time point. These differences are squared and averaged to obtain the mean square error. If the mean square error exceeds the preset range, the possible causes of excessive error are analyzed, such as unreasonable frequency band division or inaccurate filter parameter settings. Based on the analysis results, the frequency band parameters are readjusted, such as increasing or decreasing the number of frequency bands or adjusting the frequency band boundaries. The time-frequency decomposition process is repeated iteratively until the mean square error meets the preset range.

[0067] Step S300: Based on the amplitude change sequence of the instantaneous pulsation component and the time interval parameter of the corresponding sampling timestamp, construct an amplitude-time series correlation heatmap in the two-dimensional data plane. The horizontal axis of the heatmap corresponds to the time distribution of the sampling timestamp, and the vertical axis corresponds to the amplitude interval division of the instantaneous pulsation component. The gray value of the pixel in the figure is positively correlated with the frequency of pulsation at the corresponding time-amplitude coordinate.

[0068] In one implementation, step S300 may specifically include the following steps S310 to S360:

[0069] Step S310: Traverse the amplitude change sequence of the instantaneous pulsation component, count the distribution range of amplitude data in the sequence, mark the amplitude points that exceed the set interval of the distribution range as outliers, and record the sampling timestamp of the outliers.

[0070] For example, starting from the first data point of the amplitude change sequence of the instantaneous pulsation component, each amplitude data point is traversed sequentially. During the traversal, the minimum and maximum amplitude values ​​are continuously updated to obtain the distribution range of the amplitude data. Then, each amplitude data point is compared with a predefined interval of the distribution range. If an amplitude data point exceeds this interval, it is marked as an outlier. Simultaneously, the sampling timestamp corresponding to the outlier is recorded. For instance, for an amplitude change sequence of an instantaneous pulsation component, traversal reveals that the amplitude distribution range is within a preset interval, but several amplitude data points significantly exceed the predefined distribution range. These amplitude data points are marked as outliers, and their respective sampling timestamps are recorded.

[0071] Step S320: Extract the amplitude data and corresponding sampling timestamps of non-outliers, calculate the difference between the actual interval of adjacent sampling timestamps and the preset sampling interval, and generate a time interval deviation sequence. The element value of the deviation sequence is the actual interval minus the preset interval.

[0072] For example, firstly, amplitude data of non-outliers and corresponding sampling timestamps are extracted from the amplitude change sequence of instantaneous pulsations. Then, the extracted sampling timestamps are processed to calculate the actual interval between adjacent sampling timestamps. Each actual interval is subtracted from a preset sampling interval to obtain the corresponding difference. These differences are arranged in the order of sampling time to generate a time interval deviation sequence. For example, for the sampling timestamps of sequentially extracted non-outliers, the time difference between two adjacent timestamps is calculated to obtain the actual interval. Each actual interval is compared with the preset sampling interval to calculate the difference. These differences are sequentially combined into a sequence, which is the time interval deviation sequence.

[0073] Step S330: Check the time interval deviation sequence point by point. When the absolute values ​​of multiple consecutive deviation values ​​all exceed the preset deviation threshold, mark the consecutive interval as a significant deviation interval. The starting point of the interval is the timestamp of the first time that exceeds the threshold, and the ending point is the timestamp of the last time that exceeds the threshold.

[0074] In one implementation, step S330 may specifically include the following steps S331 to S336:

[0075] Step S331: Perform sliding window processing on the time interval deviation sequence. The window length is the preset number of consecutive checks. Calculate the proportion of the absolute value of the deviation value in each window that exceeds the threshold and generate a window-proportion distribution curve.

[0076] Sliding window processing involves sliding a fixed-length window across a time interval deviation sequence, analyzing the data within the window one data point at a time. The window length is a preset number of consecutive checks, which determines the number of consecutive deviation values ​​checked each time. By setting an appropriate window length, it's possible to more accurately determine whether multiple consecutive deviation values ​​exceed a threshold. The proportion of deviation values ​​exceeding the threshold within each window is calculated; this is the ratio of the number of elements with absolute deviation values ​​greater than the preset threshold to the total number of elements within the window. This proportion reflects the degree to which deviation values ​​exceed the threshold within that window. The window-proportion distribution curve is a curve with the window size on the horizontal axis and the proportion of absolute deviation values ​​exceeding the threshold on the vertical axis. By plotting this curve, the changes in the proportion of deviation values ​​exceeding the threshold at different window positions can be visually observed.

[0077] Step S332: When all the proportion values ​​in the window-proportion distribution curve are reached, mark the time interval corresponding to the window as the potentially significant deviation interval, and record the start and end timestamps of the interval.

[0078] The proportion in the window-proportion distribution curve represents the percentage of absolute deviation values ​​exceeding the threshold within each window. When the proportion reaches the maximum, it means that the absolute values ​​of all deviation values ​​within that window exceed the preset deviation threshold. This indicates that there is a severe abnormal fluctuation in the sampling time interval within the time interval corresponding to this window, and the time interval corresponding to this window is marked as a potentially significant deviation interval. Recording the start and end timestamps of the interval is to clarify the time range of this potentially significant deviation interval for further analysis and processing.

[0079] For example, the window-proportion distribution curve is checked point by point. When a window's corresponding proportion value reaches the maximum, the window's position in the time interval deviation sequence is determined. Based on the window's position, the start and end timestamps of the corresponding time interval are found and recorded.

[0080] Step S333: Merge adjacent potentially significant deviation intervals. If the end timestamps and start timestamps of two intervals are consecutive, they are merged into one interval. The start timestamp of the merged interval is the start of the previous interval, and the end timestamp is the end of the next interval.

[0081] Adjacent significant potential deviation intervals are those that appear sequentially adjacent in the time interval deviation analysis. Merging these adjacent significant potential deviation intervals helps to more accurately define the continuous time periods where significant anomalies in the sampling time intervals occur. If the end timestamps and start timestamps of two intervals are consecutive, it means that these two intervals are closely connected in time and may be caused by the same anomalous factor. Therefore, merging them into one interval can more clearly represent the range of the anomalous time period. The start timestamp of the merged interval is the start timestamp of the preceding interval, and the end timestamp is the end timestamp of the following interval, thus comprehensively covering the anomalous time range contained in the two adjacent intervals.

[0082] Step S334: Calculate the length of the significant deviation interval after merging, and determine the adjustment range of the time axis resolution based on the length. The longer the interval, the larger the adjustment range. The maximum value of the adjustment range shall not exceed the upper limit multiple of the base resolution.

[0083] For example, for each merged interval with significant deviation, its length is calculated. For instance, the interval length is obtained by subtracting the start timestamp from the end timestamp. Then, according to preset rules, the adjustment range of the time axis resolution is determined based on the interval length. A linear relationship or other functional relationship can be used to determine the adjustment range; for example, a scaling factor can be set, the interval length can be multiplied by this factor to obtain an initial adjustment range, and then this range can be compared with the upper limit multiple of the base resolution. If it exceeds the upper limit multiple, the adjustment range is set to the upper limit multiple. In this way, an appropriate time axis resolution adjustment range is determined for each merged interval with significant deviation, enabling more effective observation and handling of outlier data in subsequent analysis.

[0084] Step S335: Insert additional time scales within the significant deviation range. The number of insertions is positively correlated with the adjustment range. The inserted scales are generated by linear interpolation. The time axis is continuous, and the time scales in the non-significant deviation range maintain the basic resolution.

[0085] Inserting additional time scales within significantly biased intervals improves the resolution of the time axis within those intervals, allowing for a more detailed representation of time information and clearer observation of data changes over that period. The number of insertions is positively correlated with the adjustment magnitude; a larger adjustment magnitude results in more inserted time scales. This allows for flexible adjustment of the time axis resolution based on different anomalies. The inserted scales are generated using linear interpolation, a method of linear estimation between known data points. This method evenly inserts additional time scales between the start and end timestamps within significantly biased intervals, ensuring the continuity of the time axis. Maintaining the basic resolution time scale within non-significantly biased intervals ensures detailed analysis of these intervals while avoiding over-processing of normal time periods, thus preserving the efficiency and rationality of data processing.

[0086] Step S336: Store the adjusted time axis scale parameters as configuration information, including the start time, end time, resolution adjustment range, and insertion scale position for each interval.

[0087] The adjusted timeline scale parameters are a series of parameters obtained after adjusting the timeline resolution in areas with significant deviations. These parameters describe in detail how the timeline changes in different intervals. The configuration information includes the start time, end time, resolution adjustment range, and insertion scale position for each interval. The start and end times define the range of each interval on the timeline, the resolution adjustment range indicates the degree of change in timeline resolution within that interval, and the insertion scale position specifically indicates where additional time scales are inserted within the interval.

[0088] Step S340: Calculate the distribution density of amplitude data of non-outliers, divide the amplitude axis into several intervals, the number of intervals is positively correlated with the number of peaks in the amplitude data distribution, the interval width of dense data regions is reduced, and the interval width of sparse data regions is increased.

[0089] In one implementation, step S340 may specifically include the following steps S341 to S346:

[0090] Step S341: Statistically analyze the distribution frequency of amplitude data at non-outlier points to generate an amplitude-frequency distribution curve. The peak points on the curve correspond to the amplitude regions with dense data.

[0091] For example, the amplitude data of non-outlier points is traversed, and the frequency of each amplitude value is counted. An array or hash table can be used to store the frequency of each amplitude value. Then, the amplitudes and their corresponding frequencies are arranged in amplitude order to generate an amplitude-frequency distribution curve. For example, for a sequence of non-outlier amplitude data, the frequency of each amplitude value is counted, and the amplitude is used as the horizontal axis and the frequency of occurrence as the vertical axis to plot the amplitude-frequency distribution curve.

[0092] Step S342: Identify the peak points of the amplitude-frequency distribution curve. When the frequency of multiple consecutive amplitude points rises and then falls, mark the turning point of the trend as the peak point and record the amplitude value of the peak point.

[0093] For example, starting from the beginning of the amplitude-frequency distribution curve, examine the frequency value of each data point sequentially. When a trend of first increasing and then decreasing frequency is observed across multiple consecutive amplitude points, identify the inflection point of this trend and mark it as a peak point. Record the amplitude value corresponding to this peak point. For instance, in the amplitude-frequency distribution curve, if the frequency value gradually increases from a relatively low amplitude, reaches its maximum value at a certain amplitude point, and then begins to decrease, then this amplitude point is a peak point, and its amplitude value is recorded. By examining the entire amplitude-frequency distribution curve, identify all peak points and record their amplitude values.

[0094] Step S343: Determine the initial number of intervals for the amplitude axis based on the number of peak points. The initial number of intervals is a multiple of the number of peak points, ensuring that each peak point corresponds to at least two intervals, and the initial number of intervals is within a preset range.

[0095] The initial number of intervals for the amplitude axis is determined based on the number of peak points because peak points represent concentrated areas of data distribution, requiring sufficient intervals to describe these areas in detail. The initial number of intervals is a multiple of the number of peak points, ensuring that each peak point has enough intervals to cover its surrounding data. Ensuring that each peak point corresponds to at least two intervals allows for more accurate analysis of data changes near the peak point. Simultaneously, to avoid having too many or too few intervals, the initial number of intervals is within a preset range, which is a range pre-defined based on actual needs and data processing capabilities.

[0096] Step S344: Using the amplitude value of each peak point as the center, expand outwards to both sides to determine the initial interval boundary. The expansion distance is dynamically adjusted according to the frequency value at the peak point. The higher the frequency value, the smaller the expansion distance. The initial interval covers the dense data around the peak point.

[0097] The initial interval boundaries are determined by expanding outwards from the amplitude value of each peak point. This is to include the data around the peak point within the initial interval, allowing for more accurate analysis of data distribution. The expansion distance is dynamically adjusted based on the frequency value at the peak point. This is because a higher frequency value indicates that the data is more concentrated near that peak point, requiring less expansion distance to cover most of the data; therefore, a higher frequency value results in a smaller expansion distance. The initial interval covers the dense data around the peak point, ensuring that the data within each interval has a certain degree of correlation and representativeness.

[0098] For example, for each peak point, its amplitude value and corresponding frequency value are obtained. The expansion distance is determined based on the frequency value according to a preset rule. An inverse proportional relationship or other functional relationship can be used to determine the expansion distance; for example, a proportionality coefficient can be set, and the inverse of the frequency value can be multiplied by this coefficient to obtain the expansion distance. Then, with the amplitude value of the peak point as the center, this distance is expanded to both sides to determine the boundary of the initial interval. For example, for a peak point with a certain amplitude value and a high frequency value, a smaller expansion distance is calculated according to the rule. This distance is then expanded to both sides with this amplitude value as the center to obtain the boundary of the initial interval.

[0099] Step S345: Calculate the number of amplitude data in each initial interval, split the intervals with a data quantity exceeding a multiple of the average quantity, and the width of the split sub-intervals is half that of the original intervals; merge adjacent intervals with a data quantity lower than the average quantity, and the width of the merged intervals is the sum of the widths of the original intervals.

[0100] Calculating the number of amplitude data points within each initial interval is to understand the data distribution within each interval. The average number of amplitude data points is the average of the number of amplitude data points across all initial intervals, and this average number can be used as a reference to judge the data density of the intervals. Intervals with a data point count exceeding a multiple of the average number are split because the data in these intervals is too dense, and splitting allows for more detailed data analysis. The width of the split sub-intervals is half the width of the original intervals, which improves interval resolution while ensuring data coverage. Adjacent intervals with a data point count below the average proportion are merged because the data in these intervals is too sparse; merging reduces the number of intervals and improves data processing efficiency. The width of the merged interval is the sum of the widths of the original intervals, ensuring that the merged interval covers the range of the original two intervals.

[0101] Step S346: Smooth the boundaries of the split and merged intervals to make the width changes of adjacent intervals continuous, and generate an amplitude axis interval division scheme that includes the starting amplitude, ending amplitude and identifier of each interval.

[0102] Smoothing the boundaries of the split and merged intervals is to avoid drastic changes in interval width and to ensure continuous width changes between adjacent intervals, making the interval division of the amplitude axis more natural and reasonable. Generating an amplitude axis interval division scheme that includes the starting amplitude, ending amplitude, and identifier for each interval is to clarify the scope and identity of each interval, facilitating its use in subsequent construction of amplitude-time series correlation heatmaps.

[0103] For example, for the boundaries of the split and merged intervals, a smoothing method, such as linear interpolation or curve fitting, is used to make the width changes of adjacent intervals continuous. For instance, for two adjacent intervals, their boundary amplitudes may have large jumps. By using linear interpolation, some intermediate values ​​are inserted between the two boundary amplitudes to make the interval width changes smoother. Then, the starting amplitude, ending amplitude, and a unique identifier of each interval are compiled into a data record to generate an amplitude axis interval division scheme.

[0104] Step S350: Create a two-dimensional grid that matches the adjusted time axis and amplitude axis. Each cell of the two-dimensional grid corresponds to a time-amplitude interval. Initialize the cell count to zero. The number of rows in the grid is the same as the number of amplitude intervals, and the number of columns is the same as the number of time intervals.

[0105] Creating a two-dimensional grid that matches the adjusted time axis and amplitude axis is to integrate time and amplitude information to construct an amplitude-time series correlation heatmap. The adjusted time axis is the time axis after resolution adjustment for significantly biased intervals, and the amplitude axis is the amplitude axis after interval division. Each cell of the two-dimensional grid corresponds to a time-amplitude interval, and this grid allows for the mapping of each time point and amplitude interval combination. Initializing the cell count to zero is for subsequent counting of the number of times data appears within each time-amplitude interval. The number of rows in the grid is the same as the number of amplitude intervals, and the number of columns is the same as the number of time intervals, ensuring that the grid can completely cover all time and amplitude combinations.

[0106] Step S360: Traverse the amplitude-timestamp data pairs of non-outliers, map each data pair to the corresponding grid cell, accumulate the cell count, and add a preset mark to the count of the cell where the outlier is located. After traversal, normalize the count of each cell to the grayscale range and generate an amplitude-time series correlation heatmap.

[0107] Traversing the amplitude-timestamp data pairs of non-outliers involves processing the filtered normal amplitude data and their corresponding timestamps one by one. Mapping each data pair to a corresponding grid cell determines its cell position in the two-dimensional grid based on its amplitude and timestamp. Since each cell in the two-dimensional grid corresponds to a time-amplitude interval, the interval to which the data pair belongs can be found through the amplitude and timestamp, thus determining the corresponding cell. Accumulating the cell counts counts indicates the frequency of data occurrences within each time-amplitude interval; a higher count indicates denser data within that interval. Simultaneously, a preset marker is added to the counts of cells containing outliers to distinguish normal data from outliers in subsequent analysis; this preset marker can be a specific numerical value or symbol. After traversal, the cell counts are normalized to a grayscale range, converting the counts to grayscale values ​​to generate a visual amplitude-time series correlation heatmap.

[0108] Step S400: Perform pixel-by-pixel morphological fusion between the amplitude-time correlation heatmap and the standard pulsation pattern map stored at the time of the circuit board leaving the factory, and generate a fused difference spectrum containing the spatial distribution characteristics of the difference pixels by calculating the gray value difference of the corresponding pixels.

[0109] In one implementation, step S400 may specifically include the following steps S410 to S460:

[0110] Step S410: Read the standard pulsation pattern diagram stored at the factory of the circuit board, extract the time axis scale information, amplitude axis scale information and grayscale mapping rules of the standard pulsation pattern diagram, and perform coordinate grid resampling on the standard pulsation pattern diagram according to the spatial resolution information of the amplitude-time correlation heat map, so that the number of pixels in the time axis and amplitude axis of the resampled standard pulsation pattern diagram is the same as that of the heat map.

[0111] Reading the standard pulsation pattern diagram stored at the circuit board's factory is to obtain reference data on the pulsation pattern of the circuit board under normal conditions. The time axis scale information of the standard pulsation pattern diagram indicates the method and interval for dividing the time scale on the time axis, the amplitude axis scale information indicates the method and interval for dividing the amplitude interval on the amplitude axis, and the grayscale mapping rules specify how to convert data values ​​to grayscale values. Based on the spatial resolution information of the amplitude-time series correlation heatmap, coordinate grid resampling is performed on the standard pulsation pattern diagram to ensure that the standard pulsation pattern diagram and the amplitude-time series correlation heatmap have the same number of pixels in the time and amplitude axis directions, facilitating subsequent pixel-by-pixel comparisons. Spatial resolution information includes the number and distribution of pixels on the time and amplitude axes. The resampling process involves re-dividing and interpolating the coordinate grid of the standard pulsation pattern diagram based on the number and distribution of pixels in the amplitude-time series correlation heatmap, so that the number of pixels in the resampled standard pulsation pattern diagram in the time and amplitude axis directions is consistent with the heatmap. Figure 1 To.

[0112] Step S420: Perform pulse mode multi-resolution enhancement on the amplitude-time correlation heatmap, aggregate gray values ​​of the heatmap through different neighborhood ranges, and generate a multi-resolution heatmap set containing the original resolution and multi-level aggregated resolution. The aggregated resolution improves local gray-scale contrast by merging pixel neighborhoods.

[0113] Pulsation pattern multi-resolution enhancement is a method for processing amplitude-time series correlation heatmaps to highlight pulsation pattern features at different scales within the image. It involves aggregating grayscale values ​​in the heatmap using different neighborhood ranges. This involves selecting neighborhoods of varying sizes centered on each pixel and aggregating the grayscale values ​​of pixels within those neighborhoods, such as calculating the average or maximum value. The resulting multi-resolution heatmap set combines the original amplitude-time series correlation heatmap with the multi-resolution heatmaps obtained through neighborhood aggregation. The aggregation resolution improves local grayscale contrast by merging pixel neighborhoods. This is because during neighborhood merging, pixels with different grayscale values ​​influence each other, making local grayscale differences more pronounced and thus enhancing local grayscale contrast.

[0114] In one implementation, step S420 may specifically include the following steps S421 to S426:

[0115] Step S421: Analyze the spatial distribution of gray values ​​in the amplitude-time series correlation heatmap, and count the area ratio of areas with drastic gray value changes to areas with gradual changes. Determine the number of levels in the neighborhood range based on the area ratio. The higher the proportion of areas with drastic changes, the more levels there are.

[0116] Analyzing the spatial distribution of grayscale values ​​in an amplitude-time series correlation heatmap involves analyzing the grayscale value of each pixel in the heatmap to understand its distribution across different locations. Regions with drastic grayscale value changes refer to areas where adjacent pixels have significantly different grayscale values, potentially indicating a distinct pulsating pattern. Regions with relatively flat grayscale values ​​are those where adjacent pixels have relatively small differences in grayscale values. The area ratio of these two regions is calculated separately, representing their respective proportions within the entire heatmap. The number of neighborhood layers is determined based on these area ratios because a higher proportion of drastic changes indicates the presence of more features at different scales in the heatmap, requiring more neighborhood layers to capture these features; therefore, a higher proportion of drastic changes necessitates a greater number of neighborhood layers.

[0117] Step S422: Generate a neighborhood sequence in ascending order of neighborhood range. The neighborhood shape is rectangular. The smallest neighborhood contains a single pixel. The range of subsequent neighborhoods is gradually expanded according to the ratio of the spatial resolution of the heatmap. The maximum neighborhood range does not exceed the ratio of the shortest side of the heatmap.

[0118] Generating neighborhood sequences in ascending order of neighborhood range is for aggregating grayscale values ​​at different scales on the amplitude-time series correlation heatmap. The neighborhoods are rectangular in shape to facilitate operations and calculations on the heatmap. The smallest neighborhood contains a single pixel and serves as the starting point. Subsequent neighborhoods expand proportionally to the spatial resolution of the heatmap to ensure a regular expansion pattern that adapts to the heatmap's spatial resolution. The largest neighborhood does not exceed the proportion of the shortest side of the heatmap to prevent excessively large neighborhoods from causing the aggregation result to lose local features.

[0119] Step S423: For each neighborhood range, traverse each pixel of the heat map with the neighborhood center, calculate the average gray value of all pixels in the neighborhood as the aggregated gray value of the center pixel, and process the edge pixels through mirror filling during the aggregation process.

[0120] For each neighborhood range, each pixel in the heatmap is traversed around its center. This involves sequentially selecting each pixel on the heatmap as the neighborhood center and processing the pixels within that neighborhood. The average grayscale value of all pixels in the neighborhood is calculated as the aggregated grayscale value of the center pixel. This is done by adding the grayscale values ​​of all pixels in the neighborhood and dividing by the number of pixels in the neighborhood; the average value is then used as the aggregated grayscale value of the center pixel. Edge pixels are mirror-filled during the aggregation process because at the edges of the heatmap, the neighborhood may extend beyond the map boundary. Mirror-filling fills these extra pixels with pixels symmetrically positioned within the boundary, ensuring the integrity of the pixels within the neighborhood.

[0121] For example, for each generated neighborhood range, starting from the top-left pixel of the heatmap, each pixel is sequentially designated as the neighborhood center. For each neighborhood center, the pixels within the neighborhood are determined. If the neighborhood extends beyond the heatmap boundary, edge pixels are filled using a mirror fill method. The sum of the grayscale values ​​of all pixels within the neighborhood is calculated and then divided by the number of pixels in the neighborhood to obtain the aggregated grayscale value of the center pixel. By performing this process on each pixel of the heatmap, a resolution heatmap for that neighborhood range is obtained. Performing this process on all neighborhood ranges yields a multi-level resolution heatmap.

[0122] Step S424: Adjust the dynamic range of grayscale values ​​for each resolution heatmap, linearly mapping the grayscale values ​​of each resolution heatmap to the same grayscale range as the original heatmap, with the mapping reference being the maximum and minimum measured values ​​of the grayscale values ​​at that resolution.

[0123] Dynamically adjusting the grayscale range of heatmaps at different resolutions ensures comparability in grayscale values. Linearly mapping the grayscale values ​​of each resolution heatmap to the same grayscale range as the original heatmap involves a linear transformation to adjust the grayscale value range of each resolution heatmap to match the original range. The mapping reference is the maximum and minimum measured grayscale values ​​at that resolution; that is, the parameters for the linear mapping are determined based on the maximum and minimum grayscale values ​​in each resolution heatmap.

[0124] Step S425: Calculate the degree of grayscale difference between adjacent resolution heatmaps. The degree of difference is the average of the absolute difference of the grayscale values ​​of the corresponding pixels. If the degree of difference does not meet the preset similarity condition, remove the next resolution heatmap and retain the resolution heatmaps that meet the degree of difference condition to obtain a simplified set of multi-resolution heatmaps.

[0125] Calculating the grayscale difference between adjacent resolution heatmaps is to assess the similarity between heatmaps of different resolutions. The degree of difference is the average of the absolute differences in the grayscale values ​​of corresponding pixels. This is achieved by calculating the absolute differences in the grayscale values ​​of corresponding pixels in adjacent resolution heatmaps and then averaging these differences. If the degree of difference does not meet the preset similarity criteria, it indicates that the difference between the later resolution heatmap and the earlier resolution heatmap is small, potentially providing less additional information; therefore, the later resolution heatmap is removed. Retaining resolution heatmaps that meet the degree of difference criteria ensures that only those resolution heatmaps that provide unique information are kept, resulting in a simplified set of multi-resolution heatmaps and reducing the complexity of subsequent processing.

[0126] Step S426: Assign a weight coefficient to each resolution in the multi-resolution heatmap set. The weight coefficient is positively correlated with the degree of difference of the heatmap at that resolution. The higher the degree of difference, the higher the weight coefficient.

[0127] Assigning weight coefficients to each resolution in the multi-resolution heatmap set allows for weighted processing based on the importance of different resolution heatmaps during subsequent pulsation pattern feature point selection. The weight coefficient is positively correlated with the degree of difference between the heatmaps at that resolution; a higher degree of difference indicates more unique information provided by the heatmap, contributing more to feature point selection, hence a higher weight coefficient. For example, for each resolution heatmap in the simplified multi-resolution heatmap set, its degree of difference from adjacent resolution heatmaps is obtained. Weight coefficients are determined based on the degree of difference according to preset rules. Linear relationships or other functional relationships can be used to determine the weight coefficients; for example, a proportional coefficient can be set, and the degree of difference can be multiplied by this coefficient to obtain the weight coefficient.

[0128] Step S430: In the multi-resolution heatmap set, filter the pulse pattern feature points by the gray value gradient change characteristics. The gradient change characteristics are the ratio of the gray value of the current pixel to the average gray value of the surrounding neighborhood. Mark the pixels whose ratio meets the preset gradient conditions as candidate feature points, and record the coordinates of the candidate feature points and the corresponding gradient change characteristics at each resolution.

[0129] Filtering pulsation pattern feature points from a multi-resolution heatmap set aims to identify key pixels that represent the pulsation pattern of the circuit board. The grayscale gradient characteristic reflects the difference between the current pixel's grayscale value and the average grayscale value of its surrounding neighborhood. This characteristic helps determine whether a pixel possesses significant features. The ratio of the current pixel's grayscale value to the average grayscale value of its surrounding neighborhood is used as the gradient characteristic; a larger ratio indicates a more significant grayscale difference between the pixel and its neighbors. Pixels whose ratios meet a preset gradient condition—a pre-defined threshold—are marked as candidate feature points. Only pixels with ratios greater than this threshold are considered to have features. Recording the coordinates and corresponding gradient characteristics of candidate feature points at each resolution enables accurate location and analysis of these candidate feature points later.

[0130] In one implementation, step S430 may specifically include the following steps S431 to S436:

[0131] Step S431: For each resolution heatmap in the multi-resolution heatmap set, calculate the grayscale gradient matrix of the entire image. The value of each element in the gradient matrix is ​​the combined result of the grayscale change rate of the corresponding pixel in the horizontal and vertical directions.

[0132] Calculating the grayscale gradient matrix for each resolution heatmap in a multi-resolution heatmap set is crucial for a more comprehensive analysis of pixel grayscale variations within the heatmap. Each element in the grayscale gradient matrix represents the combined result of the corresponding pixel's grayscale change rate in both the horizontal and vertical directions. The grayscale change rate is obtained by calculating the grayscale value difference between adjacent pixels in both the horizontal and vertical directions. Combining the grayscale change rates in both directions provides a more accurate reflection of the pixel's grayscale variation trend.

[0133] Step S432: In the gradient matrix, pixels with gradient values ​​higher than the average gradient value of the heatmap at this resolution are marked as potential edge points. Non-maximum suppression is used to filter potential edge points, and local maxima points in the gradient direction are retained as edge feature points.

[0134] In the gradient matrix, pixels with gradient values ​​higher than the average gradient value of the heatmap at that resolution indicate significant grayscale changes and are likely edge locations in the image; these are marked as potential edge points. Non-maximum suppression (NMS) is a method for refining edges. This method filters potential edge points, retaining only local maxima along the gradient direction as edge feature points. The gradient direction refers to the direction of the fastest change in grayscale value; retaining local maxima along this direction makes the edges clearer and more accurate.

[0135] For example, for the gradient matrix of each resolution heatmap, the average of all elements in the matrix is ​​calculated to obtain the average gradient value of the heatmap at that resolution. Pixels with gradient values ​​higher than the average gradient value in the gradient matrix are marked as potential edge points. For each potential edge point, its gradient direction is determined. The angle of the gradient direction can be determined by calculating the ratio of the gray-level change rate in the horizontal and vertical directions. Then, along the gradient direction, the gradient values ​​of the pixel are compared with those of its neighboring pixels. If the gradient value of the pixel is greater than that of its neighboring pixels, it is retained as an edge feature point; otherwise, it is discarded.

[0136] Step S433: Delineate a circular neighborhood centered on the edge feature point, calculate the average gray value of all pixels in the neighborhood, and use the ratio of the gray value of the edge feature point to the average value of the neighborhood as the gradient change characteristic.

[0137] Delineating a circular neighborhood centered on the edge feature point allows for the analysis of pixels within a specific range around that point. Calculating the average grayscale value of all pixels within this neighborhood yields the average grayscale level of the pixels surrounding the edge feature point. The ratio of the edge feature point's grayscale value to the neighborhood average is used as a gradient characteristic. This ratio more accurately reflects the degree of grayscale difference between the edge feature point and its surrounding neighborhood, providing a more comprehensive representation of pixel characteristics than simply considering horizontal and vertical grayscale change rates.

[0138] Step S434: Obtain the preset dynamic filtering conditions for gradient change characteristics. These conditions are positively correlated with the weight coefficient of the corresponding resolution heatmap. The higher the weight coefficient, the stricter the filtering conditions. Edge feature points whose gradient change characteristics meet the corresponding filtering conditions are marked as candidate feature points.

[0139] The purpose of obtaining preset dynamic filtering conditions for gradient change characteristics is to select candidate feature points based on the importance of heatmaps at different resolutions. These conditions are positively correlated with the weight coefficients of the corresponding resolution heatmaps; higher weight coefficients indicate more important information provided by the heatmap at that resolution, requiring stricter filtering conditions to ensure the representativeness of the selected candidate feature points. Edge feature points whose gradient change characteristics meet the corresponding filtering conditions are marked as candidate feature points; only edge feature points that meet the filtering conditions are considered to have features.

[0140] For example, for each resolution heatmap, preset dynamic filtering conditions for gradient change characteristics are obtained based on its weight coefficients. Linear relationships or other functional relationships can be used to determine the filtering conditions; for example, a base threshold and a scaling factor can be set, and the filtering conditions are obtained by multiplying the weight coefficient by the scaling factor and adding it to the base threshold. The gradient change characteristics of each edge feature point are compared with the filtering conditions of the corresponding resolution heatmap. If the conditions are met, the edge feature point is marked as a candidate feature point.

[0141] Step S435: Perform spatial clustering analysis on the candidate feature points. Difference clustering is used to divide the candidate feature points that are spatially close into the same cluster. Calculate the centroid coordinates of each cluster and take the pixel point corresponding to the centroid coordinates as the representative feature point of the cluster.

[0142] Spatial clustering analysis of candidate feature points aims to group spatially adjacent candidate feature points into groups for better understanding and analysis of their distribution. Density clustering is a clustering method based on data point density, which divides spatially adjacent candidate feature points into the same cluster. The centroid coordinates of each cluster are calculated; these coordinates are the average of the coordinates of all candidate feature points in the cluster, representing the center of the cluster. The pixel corresponding to the centroid coordinates is used as the representative feature point of that cluster. This allows a single representative point to represent the entire cluster, reducing the number of feature points while preserving the main feature information of the cluster.

[0143] For example, for candidate feature points at each resolution, a density-based clustering algorithm, such as the DBSCAN algorithm, is employed. Appropriate neighborhood radii and minimum number of points are set as clustering parameters. Spatially adjacent candidate feature points are grouped into the same cluster. For each cluster, the average coordinates of all candidate feature points are calculated to obtain the centroid coordinates. The pixel corresponding to the centroid coordinates is then identified and used as the representative feature point of that cluster.

[0144] Step S436: Convert the coordinates of representative feature points at each resolution to the coordinate system of the original heatmap, remove duplicate feature point coordinates at different resolutions, retain unique feature points, obtain a cross-resolution pulsation pattern feature point set, and record the coordinates and corresponding resolution information of each feature point.

[0145] Transforming the coordinates of representative feature points at various resolutions to the coordinate system of the original heatmap is to unify feature points from different resolutions onto the scale of the original heatmap for comprehensive analysis. Since heatmaps at different resolutions may differ in pixel count and distribution, coordinate transformation is necessary. Removing duplicate feature point coordinates from different resolutions avoids redundancy, retaining only unique feature points. Retaining unique feature points yields a cross-resolution pulsation pattern feature point set, which contains representative feature points from different resolutions, providing a more comprehensive representation of the circuit board's pulsation patterns. Recording the coordinates and corresponding resolution information of each feature point allows for accurate understanding of its origin and characteristics later.

[0146] For example, for representative feature points at each resolution, coordinate transformation is performed based on the spatial resolution relationship between the heatmap at that resolution and the original heatmap. For instance, if the number of pixels in the horizontal and vertical directions of a certain resolution heatmap is half that of the original heatmap, the coordinates of the representative feature points at that resolution are multiplied by 2 to obtain their coordinates in the original heatmap's coordinate system. The transformed feature point coordinates at all resolutions are then summarized, and duplicate coordinates are removed. Data structures such as hash tables can be used for deduplication. The unique feature points retained are then used to form a cross-resolution pulsation pattern feature point set. Simultaneously, the coordinates and corresponding resolution information of each feature point are recorded.

[0147] Step S440: Perform the same multi-resolution enhancement and feature point selection process as the heatmap on the standard pulsation pattern map to generate a multi-resolution candidate feature point set for the standard pulsation pattern map, maintaining consistency between the two in terms of resolution level, neighborhood range, and gradient selection conditions.

[0148] Performing the same multi-resolution enhancement and feature point selection process on the standard pulsation pattern map as on the heatmap is to ensure comparability between the standard pulsation pattern map and the amplitude-time series correlation heatmap in terms of feature point selection. Maintaining consistency between the two in terms of resolution level, neighborhood range, and gradient selection conditions ensures that the two maps are processed under the same conditions, so that the resulting feature point sets can be directly compared and matched.

[0149] For example, the standard pulsation pattern image after resampling is processed according to the amplitude-time series correlation heatmap processing flow in steps S420-S436. First, multi-resolution enhancement of the pulsation pattern is performed, including analyzing the spatial distribution of gray values ​​to determine the number of neighborhood range levels, generating neighborhood sequences, aggregating gray values, adjusting the dynamic range of gray values, calculating the degree of gray value difference and simplifying the heatmap set, and assigning weight coefficients. Then, feature point screening is performed, including calculating the gray value gradient matrix, marking potential edge points, performing non-maximum suppression, calculating gradient change characteristics, marking candidate feature points according to dynamic screening conditions, performing spatial clustering analysis to obtain representative feature points, transforming coordinates, and removing duplicate feature points. Throughout the process, the resolution level, neighborhood range, and gradient screening conditions are maintained in relation to the amplitude-time series correlation heatmap. Figure 1 To.

[0150] Step S450: Perform bidirectional matching on the multi-resolution candidate feature points of the heatmap and the standard pulsation pattern map. Generate pixel-level coordinate mapping association by comparing the similarity of the gradient change characteristics of the feature points and the topological correspondence of their spatial locations.

[0151] Bidirectional matching of candidate feature points at multiple resolutions between the heatmap and the standard pulsation pattern map aims to identify feature points with similar characteristics in the two images. By comparing the similarity of gradient change characteristics and the topological correspondence of spatial locations, a comprehensive consideration of both feature point characteristics and location information is used to determine whether two feature points match. The similarity of gradient change characteristics reflects the consistency of feature points in grayscale changes, while the topological correspondence reflects the relative positional relationship of feature points in the images. Generating pixel-level coordinate mapping associations establishes a connection between the two images by mapping the pixel coordinates of successfully matched feature points in the two images, providing a foundation for subsequent image registration and difference analysis.

[0152] For example, for each multi-resolution candidate feature point in the heatmap, a matching point is sought in the set of multi-resolution candidate feature points in the standard pulsation pattern map. First, the similarity of gradient change characteristics between feature points is calculated, which can be measured using the absolute value of the difference or a ratio. Simultaneously, the spatial topological relationship between the two feature points is compared, evaluated by calculating their relative positions and distances in the map. A similarity threshold and a spatial position deviation threshold can be set; when the gradient change characteristic similarity is greater than the similarity threshold and the spatial position deviation is less than the spatial position deviation threshold, the two feature points are considered a match.

[0153] For each successfully matched feature point pair, their pixel coordinates in the heatmap and the standard pulsation pattern map are recorded, forming a pixel-level coordinate mapping association. By performing this matching operation on all multi-resolution candidate feature points in the heatmap, a complete pixel-level coordinate mapping association is generated.

[0154] Step S460: Based on the coordinate mapping association, the standard pulsation pattern map and the heat map are registered at the pixel level. The gray values ​​of the corresponding pixels in the registered image are calculated by difference to generate a gray-level difference matrix. The pixels in the matrix whose gray-level difference satisfies the gradient condition are marked as difference pixels. The spatial aggregation state of the difference pixels is analyzed and the connected region characteristics are extracted. The gray-level difference matrix and the connected region characteristics are fused to generate a fused difference spectrum.

[0155] Pixel-level registration of the standard pulsation pattern image and the heatmap based on coordinate mapping association is performed using the pixel-level coordinate mapping association generated in step S450 to precisely align the pixels of the standard pulsation pattern image and the heatmap. This registration ensures that the two images correspond to the same practically identical regions at the same locations, providing an accurate basis for subsequent difference analysis. A gray-level difference matrix is ​​generated by subtracting the gray-level values ​​of pixels at the same locations in the two registered images, resulting in a new matrix where each element represents the gray-level difference of the corresponding pixel. This matrix visually reflects the gray-level differences between the two images at various locations.

[0156] In the labeling matrix, pixels whose gray-level differences satisfy a gradient condition are considered difference pixels. The gradient condition is a pre-defined threshold; only pixels with gray-level differences greater than this threshold are considered to have significant differences. These pixels may represent the difference between the current operating state and the normal state of the circuit board. Analyzing the spatial clustering of difference pixels and extracting connected component characteristics involves further analysis of the regions marked as difference pixels. By analyzing the spatial distribution of difference pixels, we can identify the connected components they form and extract their characteristics, such as size, shape, and location. These characteristics can help determine the severity of the differences and their possible causes.

[0157] The grayscale difference matrix and connected region characteristics are fused to generate a fused difference spectrum. Information fusion combines the grayscale difference information contained in the grayscale difference matrix with the spatial distribution information contained in the connected region characteristics to form a more comprehensive and valuable spectrum. The fused difference spectrum can comprehensively reflect the differences in grayscale difference and spatial distribution between the current operating state and the normal state of the circuit board, providing rich information for subsequent assessment of the circuit board's operational health status.

[0158] For example, the standard pulsation pattern image and the heatmap are registered based on pixel-level coordinate mapping. Interpolation or other methods can be used to adjust the positions of pixels in the standard pulsation pattern image to align with those in the heatmap. After registration, the gray-level difference between the two images is calculated pixel-by-pixel, generating a gray-level difference matrix. The elements in the gray-level difference matrix are compared with gradient conditions, and pixels that meet the conditions are marked as difference pixels. Image segmentation algorithms, such as region growing algorithms, are used to analyze the spatial clustering of difference pixels and extract connected region characteristics. For example, using a region growing algorithm, starting from a difference pixel, adjacent difference pixels are continuously expanded to form a connected region, and the boundary, area, and other characteristics of this region are recorded. Finally, the gray-level difference matrix and connected region characteristics are fused; for example, the connected region characteristics can be added to the gray-level difference matrix using a certain encoding method to generate a fused difference spectrum.

[0159] Step S500: Calculate the state coupling coefficient of each power path based on the spatial density distribution characteristics of the difference pixels in the fusion difference spectrum, and integrate the state coupling coefficients of all power paths according to the preset weight rules to obtain a quantitative condensed index that characterizes the overall operating health status of the circuit board.

[0160] In one implementation, step S500 may specifically include the following steps S510-S560:

[0161] Step S510: Read the power path layout information of the circuit board, extract the power-on time, stable working time and power-off time of each power path, convert the stable working time into a timestamp range relative to the monitoring start time, and establish the correspondence between the power path and the heat map time interval.

[0162] The power path layout information of the circuit board is a document generated during the circuit board design process. It contains detailed information about each power path, such as its direction and the location of key nodes. It also records the power-on time, stable operating time, and power-off time for each power path. Power-on time refers to the moment the power path begins to be powered on; stable operating time refers to the period during which the power path operates normally; and power-off time refers to the moment the power path stops being powered on. Converting the stable operating time into a timestamp range relative to the monitoring start time is to unify the time information to the monitoring time scale, facilitating subsequent mapping with the time intervals of the heatmap. Establishing the correspondence between power paths and heatmap time intervals is to clarify the corresponding time range for each power path in the heatmap, enabling subsequent extraction of the difference pixels corresponding to that path.

[0163] In one implementation, step S510 may specifically include the following steps S511 to S516:

[0164] Step S511: Parse the circuit board power path layout document, extract the unique identifier, design current value and key node location of each power path, and ensure that each path is unique and distinguishable.

[0165] Analyzing the power path layout document involves detailed analysis and processing of the document to extract the necessary information. Each power path has a unique identifier, used to distinguish different power paths; this can be a number, name, etc., ensuring each path has a unique identity within the system. The design current value refers to the normal operating current specified for the power path during design; this value reflects the path's load capacity and importance. Critical node locations refer to the nodes on the power path that play a crucial role in the operation of the circuit board, such as power input / output interfaces and power supply pin connection points for major chips.

[0166] Step S512: Query the power timing control record of the circuit board to obtain the power-on time point, stable operation start time point and stable operation end time point of each power path. The timing information is expressed as the time offset relative to the monitoring start time.

[0167] Querying the power sequence control records of the circuit board is to obtain the operating sequence and status changes of each power path over time. The power-on time point refers to the moment when the power path begins to be powered on; the stable operation start time point refers to the moment when the power path reaches a stable operating state; and the stable operation end time point refers to the moment when the power path ends its stable operating state. The timing information is expressed as a time offset relative to the monitoring start time to unify the time information onto the monitoring time scale, facilitating subsequent calculations and processing.

[0168] Step S513: Convert the start and end times of the stable operation into timestamp ranges. The start timestamp is the start time of the stable operation, and the end timestamp is the end time of the stable operation. The timestamp range corresponds to the effective monitoring interval of the path.

[0169] Converting the start and end times of stable operation into a timestamp range is to represent time information in a unified and quantifiable way. The start timestamp is the time when stable operation begins, and the end timestamp is the time when stable operation ends. This timestamp range corresponds to the effective monitoring interval of the power path, meaning that within this time period, the power path is in a stable operating state and can be effectively monitored and analyzed.

[0170] Step S514: Based on the time axis configuration parameters of the heatmap, convert the timestamp range into the column index range of the heatmap. The column index start value is the column position corresponding to the start timestamp, and the column index end value is the column position corresponding to the end timestamp. The conversion is achieved through the scale mapping relationship of the time axis.

[0171] Converting timestamp ranges into column index intervals based on the heatmap's timeline configuration parameters is crucial for mapping time information to specific locations within the heatmap. This allows for the subsequent extraction of difference pixels within corresponding time intervals from the heatmap. The timeline configuration parameters include the timeline's scale mapping, specifying which column in the heatmap each timescale corresponds to. The column index start value is the column position corresponding to the start timestamp, and the column index end value is the column position corresponding to the end timestamp. This mapping accurately pinpoints the time interval of the power path within the heatmap.

[0172] Step S515: Check whether there is overlap in the column index range of each power path. If there is overlap, adjust the attribution according to the design current value of the path. The path with the preset current value obtains the main attribution record of the overlapping area and records the sharing ratio of the overlapping area.

[0173] Checking for overlap in the column index ranges of each power path ensures accurate path assignment during subsequent extraction of difference pixels. If overlapping areas exist, the assignment needs to be adjusted based on the path's design current value. Paths with higher pre-set current values ​​are typically more critical for the board's normal operation, thus granting primary assignment to overlapping areas. Recording the sharing ratio of overlapping areas ensures accurate consideration of their impact during subsequent calculations of the state coupling coefficient.

[0174] Step S516: Integrate the identification information, column index range, design current value and overlap sharing ratio of each power path to generate a power path-heat map time interval correspondence table, where each record corresponds to the time mapping information of one path.

[0175] Integrating the identification information, column index range, design current value, and overlap / sharing ratio of each power path aims to centralize all important information related to the power path and the heatmap time interval, forming a clear and complete correspondence table. The identification information uniquely distinguishes each power path, the column index range clarifies the corresponding time range in the heatmap, the design current value reflects the path's importance, and the overlap / sharing ratio considers potential overlap between paths. In the generated power path-heatmap time interval correspondence table, each record corresponds to the time mapping information of one path. This table facilitates easy querying and use of relevant information for each path, providing convenience for subsequent extraction of difference pixels and calculation of state coupling coefficients.

[0176] Step S520: Based on the correspondence, extract the difference pixels within the time interval corresponding to each power path in the fusion difference spectrum, and generate a path-difference pixel set. Each set contains the coordinates and grayscale differences of the difference pixels of the corresponding path.

[0177] Based on the power path-heatmap time interval correspondence table, the difference pixels within the corresponding time interval of each power path are extracted from the fused difference spectrum to specifically analyze the operating status of each power path. The fused difference spectrum contains information on the grayscale differences and spatial distribution between the current operating state and the normal state of the circuit board. The correspondence table allows for accurate location of the time interval corresponding to each power path in the fused difference spectrum. Extracting the difference pixels within this interval yields specific difference information related to that power path.

[0178] The generation process involves creating sets of difference pixels along a power path. Each set corresponds to a single power path and contains the coordinates and grayscale differences of the difference pixels within that path. The coordinates indicate the pixel's position within the fused difference spectrum, while the grayscale difference represents the difference in grayscale between the pixel in its current operating state and its normal operating state. This information helps analyze whether the power path exhibits any anomalies and the extent of those anomalies.

[0179] For example, for each record in the power path-heatmap time interval correspondence table, pixels within the corresponding time interval are extracted from the fused difference spectrum based on the column index interval. These pixels are compared with the labels of the difference pixels to filter out the difference pixels. The coordinates and grayscale value differences of each difference pixel are recorded and combined into a set.

[0180] Step S530: For each path-difference pixel set, calculate the ratio of the number of difference pixels to the total number of pixels in the corresponding time interval, which is used as the spatial density. Also calculate the sum of the grayscale differences of all difference pixels in the set. The spatial density and the sum of grayscale values ​​together constitute the path difference feature.

[0181] For each path-difference pixel set, the spatial density is calculated by determining the ratio of the number of difference pixels to the total number of pixels in the corresponding time interval. Spatial density reflects the density of difference pixels within the corresponding time interval of the power path; a larger ratio indicates a denser distribution of difference pixels, potentially suggesting a more severe anomaly in the power path. The sum of grayscale differences for all difference pixels in the set is also calculated. This sum represents the overall grayscale difference of the power path within the corresponding time interval; a larger sum indicates a more significant difference in grayscale compared to the normal state. Spatial density and the sum of grayscale values ​​together constitute the path difference feature, which comprehensively reflects the degree of difference between the operating state and the normal state of each power path.

[0182] Step S540: Query the historical operation database of the circuit board, extract historical records with similar difference features to the current path, and determine the similarity by the degree of closeness of the spatial density and gray sum of the path difference features. Perform a weighted average of the state coupling coefficients of similar records to generate the initial state coupling coefficient of the current path.

[0183] Querying the circuit board's historical operation database is to utilize historical data to assist in evaluating the current power path's operational status. This database stores various data from the circuit board's past operations, including path difference characteristics and state coupling coefficients. Extracting historical records similar to the current path's difference characteristics involves comparing the current path's path difference characteristics (spatial density and total grayscale) with those in the historical records to identify similar records. Similarity is determined by the closeness of the spatial density and total grayscale of the path difference characteristics. A similarity threshold can be set; records are considered similar when the differences in spatial density and total grayscale are within the threshold range.

[0184] The initial state coupling coefficient of the current path is generated by weighted averaging of the state coupling coefficients of similar historical records. The state coupling coefficient reflects the degree of interaction between the operating state of the power path and other paths. By weighted averaging the state coupling coefficients of similar historical records, a more reasonable initial state coupling coefficient can be obtained by comprehensively considering historical factors. Different weights can be assigned to the weighted average based on factors such as the time elapsed in the historical records and the reliability of the data.

[0185] For example, all records are queried from the circuit board's historical operation database, and the path difference features (spatial density and total grayscale) of each record are compared with the path difference features of the current path. The difference between the spatial density and the total grayscale is calculated; if the difference is within a similarity threshold, the record is marked as a similar record. For each similar record, its state coupling coefficient and corresponding weight are obtained. The weights can be assigned based on the time proximity of the records, with more recent records having higher weights. The state coupling coefficients of similar records are multiplied by their corresponding weights, summed, and then divided by the sum of the weights to obtain the initial state coupling coefficient of the current path.

[0186] Step S550: Analyze the electrical connection relationship of each power path, identify adjacent paths with direct connection, calculate the number of overlapping coordinates of the difference pixels of adjacent paths, and use the ratio of the number of overlapping pixels to the total number of difference pixels of the two paths as the spatial overlap. For paths with an overlap higher than a preset value, perform cross-correction of the state coupling coefficient.

[0187] Analyzing the electrical connections of each power path is to understand the physical connections between them. Directly connected adjacent paths may influence each other's operating states. Calculating the number of overlapping coordinates of differing pixels in adjacent paths quantifies the degree of this mutual influence. By comparing the coordinates of differing pixels in adjacent paths, identical coordinates are identified and their counts are recorded. The ratio of the overlapping number to the total number of differing pixels in the two paths is called the spatial overlap degree. Spatial overlap degree reflects the degree of spatial similarity between adjacent paths; a higher ratio indicates more similar differences between the two paths, and a greater potential for mutual influence.

[0188] For paths with an overlap exceeding a preset value, a cross-correction of the state coupling coefficient is performed. The preset value is a pre-defined threshold. When the spatial overlap exceeds this threshold, it indicates that the mutual influence between adjacent paths cannot be ignored, and their state coupling coefficients need to be corrected. Cross-correction refers to adjusting the state coupling coefficients of adjacent paths to more accurately reflect their mutual influence.

[0189] In one implementation, step S550 may specifically include the following steps S551 to S556:

[0190] Step S551: Based on the circuit schematic of the circuit board, extract the connection node information of each power path. When two paths are connected through the same node, they are marked as adjacent paths, and an adjacent relationship network of power paths is established.

[0191] Extracting connection node information for each power path from the circuit schematic is crucial for clarifying the physical connections between them. Connection nodes are the intersections of power paths; two paths connected by the same node are electrically connected and their operating states may influence each other. Marking adjacent paths and establishing a power path adjacency network provides a structured representation of these relationships, facilitating subsequent analysis and processing.

[0192] For example, the circuit schematic of a circuit board can be analyzed. Circuit topology analysis algorithms can be used to identify the connection nodes of each power path. For instance, using graph theory algorithms, power paths can be viewed as edges in a graph, and connection nodes as vertices. By traversing the graph, edges connecting the same vertex can be found, i.e., adjacent paths. For each pair of adjacent paths, their identification information is recorded, establishing an adjacency relationship network. This network can be represented using data structures such as adjacency matrices or adjacency lists.

[0193] Step S552: For each path in the adjacent relationship network, extract its corresponding path-difference pixel set and record the coordinates of all difference pixels in the set.

[0194] For each path in the adjacency network, extracting its corresponding path-difference pixel set is to obtain the difference pixel information of each path in the fused difference spectrum. The path-difference pixel set contains the coordinates and grayscale differences of the difference pixels for the corresponding path. Recording the coordinates of all difference pixels in the set is for subsequent calculation of the coordinate overlap between adjacent paths. For example, for each path in the adjacency network, based on the power path-heatmap time interval correspondence table, the difference pixels within the corresponding time interval of that path are extracted from the fused difference spectrum to form the path-difference pixel set. Each difference pixel in the set is traversed, and its coordinates are recorded. Data structures such as lists or dictionaries can be used to store these coordinates.

[0195] Step S553: ​​For each pair of adjacent paths, calculate the number of intersections of the coordinates of the differing pixels between the two paths. The number of intersections is the number of coordinates that appear in the set of differing pixels of both paths at the same time.

[0196] Calculating the intersection number of the differing pixel coordinates of each pair of adjacent paths quantifies the spatial similarity between them. The intersection number represents the number of coordinates that simultaneously appear in the sets of differing pixel coordinates of both paths. A higher intersection number indicates a greater spatial overlap of the differing pixel coordinates of the two paths, suggesting more similar operational states and potentially greater mutual influence. For example, for each pair of adjacent paths in an adjacency network, their lists of differing pixel coordinates are compared. The intersection number can be calculated using the intersection operation of sets.

[0197] Step S554: Calculate the ratio of the number of intersections to the total number of different pixels between the two paths. This ratio is the spatial overlap of the adjacent paths. The total number is the sum of the number of different pixels between the two paths minus the number of intersections.

[0198] The spatial overlap of adjacent paths is calculated by ratioing the number of intersections to the total number of differing pixels between the two paths. This ratio comprehensively considers both the number of differing pixels and the overlap between adjacent paths. The total number of differing pixels between the two paths is the sum of the number of differing pixels for each path minus the number of intersections. This is because pixels in the intersection are counted twice in the total calculation and need to be subtracted. Spatial overlap reflects the degree of spatial similarity between adjacent paths; a higher ratio indicates that the differences between the two paths are more similar, and the potential for mutual influence is greater.

[0199] Step S555: Obtain a preset spatial overlap threshold. When the spatial overlap of adjacent paths is higher than the threshold, start the cross correction of the state coupling coefficient. The correction amount is the initial state coupling coefficient of the adjacent paths multiplied by the spatial overlap.

[0200] A preset spatial overlap threshold is obtained. This threshold is a pre-defined critical value used to determine whether the mutual influence between adjacent paths requires correction of the state coupling coefficient. When the spatial overlap of adjacent paths exceeds the threshold, it indicates that the mutual influence between adjacent paths cannot be ignored, and cross-correction of the state coupling coefficient needs to be initiated. The correction amount is the initial state coupling coefficient of the adjacent paths multiplied by the spatial overlap. This allows the degree of correction to be adjusted according to the magnitude of the spatial overlap; the higher the spatial overlap, the greater the correction amount.

[0201] For example, a preset spatial overlap threshold is obtained from the system configuration. For each pair of adjacent paths, the calculated spatial overlap is compared with the threshold. If the spatial overlap is greater than the threshold, cross-correction of the state coupling coefficient is initiated. In this way, the state coupling coefficient of adjacent paths is reasonably corrected according to the magnitude of the spatial overlap, so that the state coupling coefficient more accurately reflects the mutual influence between adjacent paths.

[0202] Step S556: For each path, sum the corrections of all adjacent paths and add them to the initial state coupling coefficient of that path to obtain the corrected state coupling coefficient. During the correction process, ensure that the coefficient value does not exceed the preset range.

[0203] For each path, the corrections for all adjacent paths are summed to comprehensively consider the mutual influence between the current path and all its neighbors. Since a path may have multiple neighbors, each neighbor will have a certain correction effect on its state coupling coefficient. Summing these corrections yields the total correction. Adding the total correction to the initial state coupling coefficient of the path gives the corrected state coupling coefficient, which more accurately reflects the actual operating state of the path. During the correction process, it is ensured that the coefficient value does not exceed a preset range. The preset range is a pre-defined reasonable value interval for the state coupling coefficient; exceeding this range may lead to unreasonable results. If the corrected state coupling coefficient exceeds the preset range, it is adjusted to the boundary value of the preset range.

[0204] Step S560: Read the preset power path weight configuration. The weight value is determined based on the functional importance of the path in the circuit. The weight value of the core path is higher than that of the auxiliary path. Multiply the corrected state coupling coefficient of each path by the corresponding weight value, and sum them to obtain a quantitative condensed index that represents the overall operating health status of the circuit board.

[0205] The system reads the preset power path weight configuration, which is pre-set based on the functional importance of each power path in the circuit board. Core paths typically play a crucial role in the normal operation of the circuit board; for example, power paths supplying main chips have relatively high weight values. Auxiliary paths, on the other hand, play a supporting role in the normal operation of the circuit board and have relatively low weight values. Multiplying the corrected state coupling coefficient of each path by its corresponding weight value is to comprehensively consider the importance and operating status of each path. The state coupling coefficient of different paths reflects their own operational health, while the weight value reflects their importance in the entire circuit board; multiplying these two values ​​combines them.

[0206] The summation yields a quantified index characterizing the overall operational health of the circuit board. This quantified index is a comprehensive indicator that fully reflects the overall operational health of the circuit board. By adding the weighted state coupling coefficients of each path, a single value is obtained. The higher the value, the worse the overall operational health of the circuit board; the lower the value, the better the overall operational health of the circuit board.

[0207] This application provides a computer-readable storage medium storing computer-executable instructions or a computer program. When the computer-executable instructions or the computer program are executed by a processor, the processor will execute the operating status data processing method for electronic circuit boards provided in this application, for example, such as... Figure 2 The method for processing operating status data applied to electronic circuit boards is shown.

[0208] In some embodiments, the computer-readable storage medium may be a read-only memory (ROM), random access memory (RAM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), flash memory, magnetic surface memory, optical disk, or CD-ROM, etc.; or it may be a device that includes one or any combination of the above-mentioned memories.

[0209] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, and improvements made within the spirit and scope of this application are included within the scope of protection of this application.

Claims

1. A method for processing operational status data of electronic circuit boards, characterized in that, The method includes: Real-time voltage fluctuation data of each power path of the electronic circuit board is acquired within a preset monitoring period. The real-time voltage fluctuation data is synchronously acquired by high-frequency sampling units distributed at key power nodes of the circuit board at a preset sampling frequency, and includes a continuous time series of voltage amplitude change records. The real-time voltage fluctuation data is subjected to adaptive time-frequency decomposition to separate the steady-state base component, which characterizes the stable operation of the circuit board, and the instantaneous pulsation component, which reflects the instantaneous interference phenomenon. The steady-state base component is the trend fitting curve of the voltage fluctuation data on the time axis, and the instantaneous pulsation component is the difference sequence between the original data and the steady-state base component. Based on the amplitude change sequence of the instantaneous pulsation component and the time interval parameter of the corresponding sampling timestamp, an amplitude-time series correlation heatmap is constructed in a two-dimensional data plane. The horizontal axis of the heatmap corresponds to the time distribution of the sampling timestamp, and the vertical axis corresponds to the amplitude interval division of the instantaneous pulsation component. The gray value of the pixel in the graph is positively correlated with the frequency of pulsation at the corresponding time-amplitude coordinate. The amplitude-time correlation heatmap is fused with the standard pulsation pattern map stored at the time of the circuit board leaving the factory, and a fused difference spectrum containing the spatial distribution characteristics of the difference pixels is generated by calculating the gray value difference of the corresponding pixels. The state coupling coefficient of each power path is calculated based on the spatial density distribution characteristics of the difference pixels in the fused difference spectrum. The state coupling coefficients of all power paths are integrated according to a preset weighting rule to obtain a quantitative condensed index that characterizes the overall operating health status of the circuit board.

2. The method for processing operating status data of electronic circuit boards according to claim 1, characterized in that, The adaptive time-frequency decomposition of the real-time voltage fluctuation data separates the steady-state basis component characterizing the stable operation characteristics of the circuit board and the instantaneous pulsation component reflecting transient interference phenomena, including: The real-time voltage fluctuation data is scanned for time series characteristics. Different time windows are extracted according to a preset sliding step size. The fluctuation period characteristics and amplitude change rate of the voltage data in each window are extracted to generate a fluctuation characteristic spectrum containing the period-amplitude correlation. The sliding step size of the window is the same as the reciprocal of the sampling frequency. Project statistics are performed on the frequency axis of the wave characteristic spectrum, and the frequency point where the cumulative value of the amplitude change rate forms the peak value is marked. The expansion is carried out from the peak point to both sides until the cumulative value drops to the median level of the cumulative value corresponding to the peak value. The expanded interval is marked as the inherent interference frequency interval. Based on the number of inherent interference frequency intervals and the total bandwidth, determine the number of frequency bands for time-frequency decomposition, so that each interval corresponds to an independent frequency band, and the ratio of the number of frequency bands to the number of intervals is not less than a preset ratio. At the same time, insert transition frequency bands between adjacent intervals, and the bandwidth of the transition frequency band is the average of the bandwidths of the adjacent intervals. Real-time voltage fluctuation data is input into the time-frequency decomposition process according to the determined frequency band parameters. The data is divided into frequency bands with multi-resolution resolution to generate energy distribution matrices for different frequency bands. The row dimension of the matrix corresponds to the time axis, the column dimension corresponds to the frequency axis, and the element value represents the energy density of the corresponding time-frequency point. A preset low-frequency band is defined on the frequency axis of the energy distribution matrix. Within this band, regions with energy density higher than the average level of adjacent regions and fluctuation amplitude less than a preset fluctuation threshold within multiple consecutive time windows are identified and marked as the base signal frequency band. The energy distribution of the base signal frequency band is inversely transformed to obtain the preliminary extraction results of the steady-state base component. A preset high-frequency band is defined on the frequency axis. Regions with energy density change rate exceeding the average change rate of adjacent time windows within a single time window are identified and marked as the interference signal frequency band. The energy distribution of the interference signal frequency band is denoised to obtain the preliminary extraction results of the instantaneous pulsation component. The preliminary results of the steady-state base component and the preliminary results of the instantaneous pulsation component are superimposed in the time domain. The superimposed result is compared with the original real-time voltage fluctuation data, and the mean square error of the two is calculated. If the error exceeds the preset range, the frequency band parameters are readjusted and the decomposition process is repeated until the error meets the requirements, thus generating the final steady-state base component and instantaneous pulsation component.

3. The method for processing operating status data of electronic circuit boards according to claim 2, characterized in that, The frequency axis of the wave characteristic spectrum is projected and statistically analyzed to mark the frequency point where the cumulative value of the amplitude change rate forms a peak. The frequency range is then expanded outwards from the peak point until the cumulative value drops to the median level corresponding to the peak value. This expanded range is marked as the inherent interference frequency range, including: The frequency axis of the wave characteristic spectrum is scanned point by point, and the cumulative value of the amplitude change rate corresponding to each frequency point is counted to generate a frequency-cumulative value distribution curve. The peak point of the curve corresponds to the potential interference frequency center. Traverse the frequency-cumulative value distribution curve. When the cumulative value of multiple consecutive frequency points shows a trend of first rising and then falling, mark the trend inflection point as the peak point and record the frequency value and corresponding cumulative value of the peak point. Starting with the frequency value of each peak point, the frequency-cumulative value distribution curve is scanned point by point towards the lower frequency direction until the cumulative value drops to the median level of the peak cumulative value. The frequency value at this position is recorded as the low-frequency boundary. The same scan is performed towards the higher frequency direction, and the high-frequency boundary is recorded. The frequency range between the low-frequency boundary and the high-frequency boundary is marked as the candidate interference frequency interval. Each peak point corresponds to a candidate interval, and the range of the candidate interval is defined by the low-frequency boundary and the high-frequency boundary. Read the normal interference bandwidth range from the circuit board design document, compare the bandwidth of the candidate interference frequency interval with the normal range, and mark the candidate interval with the bandwidth exceeding the normal range as the inherent interference frequency interval. The inherent interference frequency ranges marked are checked for overlap. If the boundaries of adjacent ranges overlap, the overlapping parts are merged into a single range. The boundaries of the merged range are the minimum low-frequency boundary and the maximum high-frequency boundary of the original range, generating the final set of inherent interference frequency ranges.

4. The method for processing operating status data of electronic circuit boards according to claim 3, characterized in that, The step of inputting real-time voltage fluctuation data into a time-frequency decomposition process according to determined frequency band parameters, performing multi-resolution frequency band division on the data, and generating energy distribution matrices for different frequency bands includes: Based on the bandwidth of each interval in the inherent interference frequency interval set and the historical fault association records, the dynamic influence factor of each interval is calculated. The larger the bandwidth and the more frequently the interval appears in the historical fault association records, the larger the influence factor. The real-time voltage fluctuation data is then subjected to time-domain segmentation enhancement processing through the interval influence factor. Based on the number and frequency distribution characteristics of inherent interference frequency intervals, a multi-resolution decomposition frequency band tree structure is constructed. The hierarchical depth of the tree structure is positively correlated with the number of intervals. Each level corresponds to a frequency band. The frequency band boundaries between levels are determined by the frequency axis bisection method of the previous level, so that the frequency band division covers all inherent interference frequency intervals. For each level of the tree structure, the filter parameters of that level are dynamically configured according to the start frequency, end frequency and isolation requirements of the corresponding frequency band. The passband range of the filter matches the frequency boundary of the frequency band, and the stopband range avoids the frequency range of the adjacent frequency band, so as to perform frequency band isolation processing on the real-time voltage fluctuation data. For voltage data that has undergone frequency band isolation processing, signal components are extracted sequentially according to hierarchical order. Each extracted signal component corresponds to an inherent interference frequency range. During the extraction process, the crosstalk between the current frequency band and the frequency bands above and below is monitored in real time. The crosstalk is calculated by the ratio of signal power leakage to the signal power of the current frequency band. When the crosstalk exceeds the preset threshold, the transition band slope of the current level filter is adjusted. The transition band slope is positively correlated with the crosstalk. At the same time, the frequency band boundary of the previous level is backtracked and optimized, and the isolation band width of adjacent frequency bands is expanded or reduced until the crosstalk drops below the threshold. For the signal components extracted at each level, the square of the signal amplitude at each sampling point is calculated in time order as the energy density value. The energy density values ​​of different levels are arranged in frequency band order to obtain a two-dimensional matrix with time as the row and frequency band as the column. The matrix element values ​​reflect the degree of energy concentration at the corresponding time-frequency band point.

5. The method for processing operating status data of electronic circuit boards according to claim 1, characterized in that, The method of constructing an amplitude-time series correlation heatmap in a two-dimensional data plane based on the amplitude change sequence of the instantaneous pulsation component and the time interval parameter of the corresponding sampling timestamp includes: Traverse the amplitude change sequence of instantaneous pulsation components, count the distribution range of amplitude data in the sequence, mark amplitude points that exceed the set interval of the distribution range as outliers, and record the sampling timestamp of the outliers; Extract the amplitude data and corresponding sampling timestamps of non-outliers, calculate the difference between the actual interval of adjacent sampling timestamps and the preset sampling interval, and generate a time interval deviation sequence. The element value of the deviation sequence is the actual interval minus the preset interval. The time interval deviation sequence is checked point by point. When the absolute values ​​of multiple consecutive deviation values ​​exceed the preset deviation threshold, the continuous interval corresponding to the absolute values ​​of the multiple consecutive deviation values ​​is marked as the deviation significant interval. The starting point of the interval is the timestamp of the first time that exceeds the threshold, and the ending point is the timestamp of the last time that exceeds the threshold. The distribution density of amplitude data of non-outliers is statistically analyzed, and the amplitude axis is divided into several intervals. The number of intervals is positively correlated with the number of peaks in the amplitude data distribution. The interval width of dense data regions is reduced, while the interval width of sparse data regions is increased. Create and adjust a two-dimensional grid that matches the time axis and amplitude axis. Each cell of the two-dimensional grid corresponds to a time-amplitude interval. Initialize the cell count to zero. The number of rows in the grid is the same as the number of amplitude intervals, and the number of columns is the same as the number of time intervals. Iterate through the amplitude-timestamp data pairs of non-outliers, map each data pair to the corresponding grid cell, accumulate the cell counts, and add a preset mark to the count of the cell where the outlier is located. After the iteration is completed, normalize the counts of each cell to the grayscale range and generate an amplitude-time series correlation heatmap.

6. The method for processing operating status data of electronic circuit boards according to claim 5, characterized in that, The step of checking the time interval deviation sequence point by point, when the absolute values ​​of multiple consecutive deviation values ​​all exceed a preset deviation threshold, marks the continuous interval corresponding to the absolute values ​​of these multiple consecutive deviation values ​​as a significant deviation interval. The starting point of the interval is the timestamp of the first time the threshold is exceeded, and the ending point is the timestamp of the last time the threshold is exceeded. The time interval deviation sequence is processed by a sliding window, with the window length being the preset number of consecutive checks. The proportion of the absolute value of the deviation value in each window that exceeds the threshold is calculated, and a window-proportion distribution curve is generated. When all the proportion values ​​in the window-proportion distribution curve are reached, the time interval corresponding to the window is marked as a potentially significant deviation interval, and the start and end timestamps of the interval are recorded. Adjacent intervals with significant potential deviations are merged. If the end timestamps and start timestamps of two intervals are consecutive, they are merged into one interval. The start timestamp of the merged interval is the start of the previous interval, and the end timestamp is the end of the next interval. Calculate the length of the significant deviation interval after merging, and determine the adjustment range of the time axis resolution based on the length. The longer the interval, the larger the adjustment range. The maximum value of the adjustment range shall not exceed the upper limit multiple of the base resolution. Additional time scales are inserted in the significant deviation range, with the number of insertions being positively correlated with the adjustment magnitude. The inserted scales are generated through linear interpolation, ensuring a continuous time axis. The time scales in the non-significant deviation range maintain the basic resolution. The adjusted time axis scale parameters are stored as configuration information, including the start time, end time, resolution adjustment range, and insertion position of each interval.

7. The method for processing operating status data of electronic circuit boards according to claim 5, characterized in that, The distribution density of the statistical non-outlier amplitude data is determined by dividing the amplitude axis into several intervals. The number of intervals is positively correlated with the number of peak values ​​in the amplitude data distribution. The interval width decreases in densely populated regions and increases in sparsely populated regions, including: The distribution frequency of amplitude data at non-outlier points is statistically analyzed to generate an amplitude-frequency distribution curve. The peak points on the curve correspond to the amplitude regions with dense data. Identify the peak points of the amplitude-frequency distribution curve. When the frequency of multiple consecutive amplitude points first increases and then decreases, mark the turning point of the trend as the peak point and record the amplitude value of the peak point. The initial number of intervals for the amplitude axis is determined based on the number of peak points. The initial number of intervals is a multiple of the number of peak points, ensuring that each peak point corresponds to at least two intervals, and the initial number of intervals is within a preset range. The initial interval boundary is determined by expanding outwards from the amplitude value of each peak point. The expansion distance is dynamically adjusted according to the frequency value at the peak point. The higher the frequency value, the smaller the expansion distance. The initial interval covers the dense data around the peak point. Calculate the number of amplitude data in each initial interval. Divide the intervals with a number of data exceeding a multiple of the average number, and the width of the sub-intervals after the division is half that of the original intervals. Merge adjacent intervals with a number of data below the average number, and the width of the merged interval is the sum of the widths of the original intervals. The boundaries of the split and merged intervals are smoothed to ensure that the width of adjacent intervals changes continuously, generating an amplitude axis interval division scheme that includes the starting amplitude, ending amplitude, and identifier of each interval.

8. The method for processing operating status data of electronic circuit boards according to claim 1, characterized in that, The step of performing pixel-by-pixel morphological fusion of the amplitude-time correlation heatmap and the standard pulsation pattern map stored at the time of circuit board manufacturing, and generating a fused difference spectrum containing the spatial distribution features of the differing pixels by calculating the gray value differences of the corresponding pixels, includes: Read the standard pulsation pattern diagram stored at the circuit board factory, extract the time axis scale information, amplitude axis scale information and grayscale mapping rules of the standard pulsation pattern diagram, and perform coordinate grid resampling on the standard pulsation pattern diagram according to the spatial resolution information of the amplitude-time correlation heat map, so that the number of pixels in the time axis and amplitude axis of the resampled standard pulsation pattern diagram is the same as that of the heat map. Perform pulsed mode multi-resolution enhancement on the amplitude-time correlation heatmap, aggregate grayscale values ​​of the heatmap through different neighborhood ranges, and generate a multi-resolution heatmap set containing the original resolution and multi-level aggregated resolution. The aggregated resolution improves local grayscale contrast by merging pixel neighborhoods. In the multi-resolution heatmap set, the pulse pattern feature points are screened by the gray value gradient change characteristics. The gradient change characteristics are the ratio of the gray value of the current pixel to the average gray value of the surrounding neighborhood. Pixels whose ratios meet the preset gradient conditions are marked as candidate feature points. The coordinates of the candidate feature points and the corresponding gradient change characteristics are recorded at each resolution. Perform the same multi-resolution enhancement and feature point selection process as the heatmap on the standard pulsation pattern map to generate a multi-resolution candidate feature point set for the standard pulsation pattern map; A bidirectional matching process is performed on the candidate feature points of the heat map and the standard pulsation pattern map at multiple resolutions. By comparing the similarity of the gradient change characteristics of the feature points and the topological correspondence of their spatial locations, a pixel-level coordinate mapping association is generated. Based on coordinate mapping, the standard pulsation pattern map and the heat map are registered at the pixel level. The gray values ​​of corresponding pixels in the registered image are then interpolated to generate a gray-level difference matrix. Pixels in the matrix whose gray-level differences satisfy the gradient condition are marked as difference pixels. The spatial aggregation state of the difference pixels is analyzed and the characteristics of connected regions are extracted. The gray-level difference matrix and the characteristics of connected regions are then fused to generate a fused difference spectrum.

9. The method for processing operating status data of electronic circuit boards according to claim 1, characterized in that, The process involves calculating the state coupling coefficient of each power path based on the spatial density distribution characteristics of the difference pixels in the fused difference spectrum, integrating the state coupling coefficients of all power paths according to a preset weighting rule, and obtaining a quantitative condensed index characterizing the overall operational health status of the circuit board, including: Read the power path layout information of the circuit board, extract the power-on time, stable working time and power-off time of each power path, convert the stable working time into a timestamp range relative to the monitoring start time, and establish the correspondence between the power path and the heat map time interval. Based on the correspondence, the difference pixels within the time interval corresponding to each power path are extracted from the fused difference spectrum to generate a path-difference pixel set. Each set contains the coordinates and grayscale differences of the difference pixels of the corresponding path. For each path-difference pixel set, calculate the ratio of the number of difference pixels to the total number of pixels in the corresponding time interval, which is used as the spatial density; and calculate the sum of gray value differences of all difference pixels in the set. The spatial density and the sum of gray values ​​together constitute the path difference feature. The system queries the historical operation database of the circuit board and extracts historical records with similar differences to the current path. The similarity is determined by the degree of closeness of the spatial density and gray sum of the path difference features. The state coupling coefficients of similar records are weighted and averaged to generate the initial state coupling coefficient of the current path. The electrical connection relationship of each power path is analyzed, adjacent paths with direct connection are identified, the number of overlapping coordinates of the difference pixels of adjacent paths is calculated, and the ratio of the number of overlapping pixels to the total number of difference pixels of the two paths is used as the spatial overlap. Paths with an overlap higher than the preset value are cross-corrected for the state coupling coefficient. The preset power path weight configuration is read. The weight value is determined based on the functional importance of the path in the circuit. The core path has a higher weight value than the auxiliary path. The corrected state coupling coefficient of each path is multiplied by the corresponding weight value, and the sum is obtained to obtain a quantitative condensed index that represents the overall operating health status of the circuit board.

10. A computer system, characterized in that, include: Memory is used to store executable instructions or computer programs. A processor, when executing computer-executable instructions or computer programs stored in the memory, implements the operating status data processing method for an electronic circuit board as described in any one of claims 1 to 9.

Images