Signal detection assembly and noise detection method for circuit board

Abstract

The invention discloses a signal detection assembly and a noise detection method for a circuit board, and belongs to the technical field of circuit detection. The method comprises the following steps: sending a power-on signal to enable a circuit board to enter a power-on state; sound collection is carried out on the circuit board in the power-on state, and an audio signal is obtained; analyzing the audio signal based on an AI model, and identifying and determining a noise area on the circuit board; and performing waveform detection on the circuit node corresponding to the noise area, determining the position of a noise source, and outputting a detection result containing the position information of the noise source. Through combination of acoustic acquisition and an artificial intelligence algorithm, automatic identification and positioning of implicit noise such as high-frequency howling and abnormal vibration of the circuit board are realized, and compared with a traditional artificial listening and oscilloscope point-by-point detection mode, the detection efficiency and the positioning precision can be remarkably improved, the manual judgment error is reduced, and the detection accuracy is improved. The method is suitable for electronic product production detection and fault diagnosis scenes.

Description

Technical Field

[0001] This application relates to the field of circuit testing technology, and more particularly to a signal detection component and a noise detection method for a circuit board. Background Technology

[0002] Printed circuit board assemblies (PCBAs) are widely used in various electronic products, integrating a variety of components such as inductors, capacitors, transformers, and crystal oscillators. During PCBA operation, factors such as unreasonable circuit design, improper component selection, excessive power supply ripple, or electromagnetic coupling interference often lead to phenomena such as inductor howling, noise generated by the piezoelectric effect of ceramic capacitors, and circuit self-oscillation. This type of noise not only affects product quality and user experience but may also lead to rework, customer complaints, and even damage to brand reputation.

[0003] In existing technologies, PCBA noise detection mainly relies on manual listening or the use of a single detection method in an experimental environment. Examples include sound acquisition using a single-point microphone, detection of hot spots using infrared thermal imaging, or measurement of current and voltage signals using an oscilloscope to determine anomalies. However, these methods generally suffer from the following problems: Inaccurate positioning: It is difficult for human listening to accurately determine the location of noise sources in noisy environments and densely packed components; a single-point microphone can only capture the overall noise and cannot distinguish specific components.

[0004] The detection methods are limited and prone to misjudgment: thermal imaging can only identify abnormal temperature rise, and some components that produce howling may not be accompanied by significant heat generation, leading to missed detection; although oscilloscopes can display abnormal frequencies, they are difficult to directly correspond to specific sound source devices.

[0005] Lack of intelligent and systematic detection processes: Existing methods rely on experience, have low detection efficiency, poor repeatability of results, and make it difficult to form a closed loop from detection to localization and then to correction.

[0006] Therefore, how to accurately locate the position of noise source components and output clear test results by combining multiple detection methods while the PCBA is powered on has become an urgent technical problem to be solved. Summary of the Invention

[0007] The main objective of this invention is to provide a noise detection method for circuit boards. This method utilizes the collaborative work of a sound acquisition array, an oscilloscope, a thermal imager, and a terminal to detect and locate noise generated by the circuit board during power-on operation. The invention employs a sound acquisition array to collect audio signals from different locations on the circuit board, and then uses an AI model on the terminal to analyze the audio signals and identify noise regions. Subsequently, the oscilloscope performs spectral analysis of the electrical signals, and the thermal imager detects temperature distribution to cross-validate the noise regions, thereby accurately locating the noise source components.

[0008] To achieve the above objectives, the present invention provides a noise detection method for a circuit board, the method comprising: Send a power-on signal to put the circuit board into a powered state; After entering the power-on state, sound is collected from the circuit board in the power-on state to obtain an audio signal; The audio signal is analyzed using an AI model to identify the noise region; Waveform detection is performed on the circuit nodes in the noise region to locate the noise source, and the detection result including the location information of the noise source is output.

[0009] In one embodiment, the audio signal is a multi-channel audio signal.

[0010] In one embodiment, the analysis of the audio signal based on an AI model to obtain the noise region includes: The audio signal is input into the AI ​​model; The arrival time difference between each channel of the audio signal is calculated, and the coarse localization region of the noise source is obtained based on the arrival time difference; Within the coarse positioning area, delay compensation and superposition are performed on audio signals from different directions to obtain a precise positioning area corresponding to the power peak.

[0011] In one embodiment, analyzing the audio signal based on an AI model to obtain the noise region further includes: A mapping relationship is established between the precise positioning area and the circuit board layout diagram, and the confidence level of the mapping relationship is verified to obtain the confidence level. When the confidence level is greater than a preset first threshold, the precise positioning area is determined to be a noise area.

[0012] In one embodiment, using an oscilloscope to perform waveform detection on the circuit nodes in the noise region to locate the noise source includes: In the circuit nodes corresponding to the noise region, electrical signals are collected, and the signal characteristics of the electrical signals are analyzed to form a multi-dimensional parameter set. The multi-dimensional parameter set is input into the AI ​​model for processing and matched with a preset noise fault waveform database. When the matching result is higher than a preset second threshold, the corresponding circuit node is identified as a noise source.

[0013] In one embodiment, the step of performing feature analysis on the signal characteristics of the electrical signal to form a multi-dimensional parameter set includes: Extracting frequency domain features, time domain features, and statistical features of electrical signals; The characteristic parameters of the frequency domain features include at least harmonic peak value and spectral flatness; the characteristic parameters of the time domain features include at least rising edge jitter, amplitude overshoot and periodic noise pulse; and the characteristic parameters of the statistical features include at least standard deviation and peak value variation. All feature parameters are constructed into a multi-dimensional parameter set.

[0014] In one embodiment, the method further includes: Before performing the waveform detection, the noise area is scanned with an infrared sensor to obtain the temperature distribution map of the area and identify abnormal hot spots. The abnormal hot spot is used as auxiliary information and fused with the noise area obtained by the audio acquisition array for analysis. When the abnormal hot spot coincides with the noise area or is close to the preset third threshold, the confidence of the noise source determination is improved. If the abnormal heat points do not appear significantly, the weight of the thermal imaging results is reduced.

[0015] In addition, to achieve the above objectives, the present invention also provides a signal detection component, which includes: a sound acquisition array, an oscilloscope, a thermal imager, and a terminal; The sound acquisition array, oscilloscope, and thermal imager all establish a signal transmission relationship with the terminal; The sound acquisition array includes multiple miniature microphone units distributed according to a preset array position relationship, used to acquire audio signals from different positions on the circuit board; The terminal includes a processor and a memory, the processor being configured to use an AI model and an AI model to determine the noise region of the sound source on the circuit board; The oscilloscope includes a probe unit and a signal processing unit. The probe unit is used to connect to the circuit node corresponding to the noise area. The signal processing unit is used to perform spectrum and timing analysis on the acquired voltage and current signals, and match them with a preset fault waveform database based on an artificial intelligence model to identify the noise source. The thermal imager includes an infrared sensing unit and an imaging processing unit, which are used to perform infrared scanning on the noise area, generate a temperature distribution map, and fuse abnormal hot spots as auxiliary information with the acoustic localization results to improve the confidence of noise source identification.

[0016] In one embodiment, the sound acquisition array includes: Multiple miniature microphone units are distributed according to a preset array position relationship. The spatial position of each microphone unit relative to the array reference point is pre-calibrated to synchronously collect audio signals from different positions when the circuit board is powered on. The front-end analog conditioning module is connected to the output of each microphone unit and is used to amplify, filter, and adjust the gain of the audio signal with low noise. The analog-to-digital conversion and clock synchronization modules are connected to the front-end analog conditioning module and are used to synchronously sample the audio signals of each channel and output digital signals with timestamps. A shielding and vibration damping structure is installed around the sound acquisition array to suppress environmental mechanical vibration and electromagnetic interference; The data interface module is used to transmit the multi-channel digital signals to the terminal in the form of a high-speed data bus for the purpose of performing sound source localization.

[0017] In one embodiment, the terminal includes: A processor and a memory, wherein the memory is used to store artificial intelligence algorithm programs and a preset noise fault database; The processor is configured to: The system receives audio signals from a sound acquisition array, performs time difference estimation on the audio signals, and obtains a coarse localization area of ​​the noise source. Based on the coarse positioning area, delay compensation and superposition are performed on audio signals from different directions to obtain the precise positioning area corresponding to the power peak. The AI ​​model is invoked to verify the confidence level of the precisely located area, and the noise source device is determined by combining the circuit board layout diagram; The system receives detection data from an oscilloscope and a thermal imager, performs fusion analysis on the detection data and acoustic detection results, and outputs detection results including noise source location information based on the fusion results. Attached Figure Description

[0018] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of the detection method according to an embodiment of the present invention; Figure 2 This is a schematic flowchart of a method for determining noise regions using an AI model according to another embodiment of the present invention; Figure 3 This is a schematic diagram of the architecture of a signal detection component according to another embodiment of the present invention.

[0021] Explanation of icon numbers:

[0022] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Well-known modules, units, and their connections, links, communications, or operations are not shown or described in detail. Furthermore, the described features, architectures, or functions can be combined in any way in one or more embodiments. Those skilled in the art should understand that the various embodiments described below are only for illustrative purposes and are not intended to limit the scope of protection of the present invention.

[0024] Because existing circuit board (PCBA) noise detection typically relies on manual listening or a single detection method.

[0025] Manual listening is difficult to accurately distinguish specific noise sources in noisy environments or on circuit boards with dense components, and the test results are greatly affected by subjective factors.

[0026] When using an oscilloscope or thermal imager for auxiliary detection, the oscilloscope can only reflect local electrical signal waveform changes, which is difficult to correspond one-to-one with the location of the sound source. While thermal imagers can identify abnormal heating, they are less effective at detecting howling noises that do not show a significant temperature rise.

[0027] Furthermore, existing technologies lack a unified analysis mechanism for multiple test results, making it impossible to achieve intelligent localization and comprehensive judgment of circuit board noise, resulting in low testing efficiency, insufficient accuracy, and high rework rate.

[0028] This application provides a noise detection method for circuit boards and a corresponding signal detection component.

[0029] This solution enables multi-dimensional detection and intelligent judgment of circuit board noise through the coordinated operation of a sound acquisition array, oscilloscope, thermal imager, and terminal.

[0030] Specifically, the sound acquisition array is arranged around the circuit board to acquire multi-channel audio signals; the terminal analyzes the acquired audio signals to determine the noise area of ​​the sound source on the circuit board.

[0031] The oscilloscope acquires electrical signals within the noise region, extracts and analyzes the frequency domain, time domain, and statistical characteristics of the electrical signals to form a multi-dimensional parameter set, and compares it with a preset noise fault database to identify the circuit nodes that generate noise.

[0032] Before waveform detection, a thermal imager can be used to perform infrared scanning on the noise area to obtain a temperature distribution map and identify abnormal hot spots. This information can be used as auxiliary information to fuse and analyze with the acoustic localization results, thereby improving the confidence level of noise source determination.

[0033] The terminal integrates acoustic, electrical, and thermal detection results and outputs detection results including noise source location information, enabling accurate localization and automated identification of circuit board noise.

[0034] It should be noted that some technical terms mentioned in this invention have, but are not limited to, the following meanings: A "noise region" refers to a spatial area in the plane coordinate system of a circuit board where the sound pressure level or sound energy density is significantly higher than the background noise threshold within a specific frequency band, as determined by acoustic characteristic analysis of multi-channel audio signals. This region represents the local area where a suspected noise source is located and can be represented by a continuous high-energy region in a sound pressure distribution map or sound energy distribution map.

[0035] The "coarse localization area" refers to the preliminary localization range of the sound source calculated based on the time difference of arrival (TDOA) of the multi-channel audio signals. This area is usually a geometric intersection formed by the intersection of multiple time difference equations, representing the approximate location of the noise source, with the localization error preferably within the centimeter range.

[0036] "Precise positioning area" refers to the spatial range determined by selecting the power peak and its neighborhood after delay compensation and superposition of signals from different directions to form a power distribution, based on the coarse positioning area. This area is used to represent the high-confidence location of the noise source obtained after focusing calculation, and the positioning accuracy can usually reach the sub-centimeter level.

[0037] "Confidence score" refers to the score output by the AI ​​algorithm or AI model used to characterize the credibility of the detection result, with a value ranging from 0 to 1. The higher the value, the higher the credibility of the model for the current location or recognition result. When the confidence score is higher than a preset threshold, the system will determine the corresponding area or node as a valid noise source.

[0038] "Preset first threshold" refers to the threshold parameter used to determine whether the confidence level meets the conditions for identifying the noise source. It is usually initialized by the system or set by the user, and the value can be adjusted between 0.7 and 0.99.

[0039] A "multi-dimensional parameter set" refers to a combined data structure of feature parameters extracted from multiple dimensions, including the time domain, frequency domain, and statistical domain, in electrical signal feature analysis. This set may include, but is not limited to, harmonic peak values, spectral flatness, rising edge jitter, amplitude overshoot, periodic pulse frequency, standard deviation, and peak-to-peak value variation, and is used to comprehensively describe the noise characteristics of electrical signals.

[0040] "Similarity matching" refers to the process of comparing the multi-dimensional parameter set of the electrical signal under test with samples in a pre-set noise fault waveform database to calculate their similarity. Similarity can be achieved through cosine similarity, correlation coefficient, or other normalized distance metrics, with values ​​ranging from 0 to 1.

[0041] The "preset second threshold" refers to a threshold parameter used to determine whether the similarity matching result indicates the presence of a noise fault. When the similarity value is greater than this threshold, the system determines that the current node is a noise source node. The threshold range can be between 0.7 and 0.95.

[0042] The "preset third threshold" refers to a threshold used to measure the difference in distance or temperature rise between an abnormal heat source in thermal imaging and an acoustic noise area. It may include a spatial distance threshold (e.g., ≤5mm) and a temperature difference threshold (e.g., ≥0.5℃). When the distance between the abnormal heat source and the noise area is less than this threshold or the temperature rise is greater than the threshold, the two are considered correlated.

[0043] "Fusion analysis" refers to the process of weighting or fusing multi-source data from acoustic testing, oscilloscope testing, and thermal imaging testing at a terminal to obtain a comprehensive judgment result. This can include rule-based weighted fusion, confidence-based weighted averaging, or feature-level fusion using artificial intelligence models.

[0044] "Signal transmission relationship" refers to the data communication connection established between the sound acquisition array, oscilloscope, thermal imager and terminal. It can be a wired connection (including but not limited to USB, Ethernet) or a wireless connection (including but not limited to Wi-Fi, Bluetooth) to realize data acquisition and transmission.

[0045] "Preset array positional relationship" refers to the geometric arrangement and relative positional parameters of each microphone unit in the sound acquisition array in three-dimensional space, including coordinates, spacing and azimuth angle, which have been determined by calibration before the system is used.

[0046] "Synchronous sampling" or "clock synchronization" means that the sampling operations of each channel of the sound acquisition array are carried out under the unified clock control, and the time deviation between channels is less than one-tenth of the sampling period, so as to ensure the time consistency of the sampled signal.

[0047] "High-speed data bus" refers to a data interface used to transmit multi-channel digital audio signals between the sound acquisition array and the terminal. The transmission rate is preferably greater than 100 Mbps, and it can adopt standard bus forms such as USB, Ethernet or PCIe.

[0048] "Weight" refers to the relative influence coefficient assigned to different detection methods (acoustic, oscilloscope, thermal imaging) in multi-source data fusion, used to control the contribution of each detection result in the final judgment. The weight can be dynamically adjusted according to the confidence level; when a detection result is unreliable, its weight is automatically reduced.

[0049] "Power-on signal" refers to the control signal that turns on the power supply of the circuit board and puts it into working state. It is usually issued by the test system or terminal.

[0050] "Power-on status" means that all power lines on the circuit board are energized and the electrical components are in normal or test operation status.

[0051] "Audio signal" refers to a processable digital signal obtained by amplifying and converting sound waves collected by a microphone unit into digital signals, which may include single-channel or multi-channel signals.

[0052] "Multi-channel audio signal" refers to a collection of audio data acquired synchronously by multiple independent sampling channels. Each channel corresponds to a different acquisition location and is used to calculate the direction or location of the sound source.

[0053] "Time difference of arrival" refers to the time difference between when the same sound source signal is received at different microphone units. This time difference reflects the relative spatial relationship between the sound source and each microphone, and can be used to calculate the location of the sound source.

[0054] "Delay compensation" refers to applying a corresponding time delay to the signals of each channel according to the possible direction or location of the sound source, so that the signals from the target direction are aligned on the time axis.

[0055] "Superposition" refers to adding (or superimposing) the delayed-compensated signals to enhance the signal in the target direction and suppress interference from other directions. This process is a fundamental operation of beamforming algorithms.

[0056] "Power peak" refers to the point where the power spectral density of the signal reaches its maximum value after delay superposition, corresponding to the location or direction where the sound source energy is strongest.

[0057] An oscilloscope is a test instrument used to acquire, display, and analyze changes in electrical signals such as voltage and current over time. It can include digital oscilloscopes or mixed-signal oscilloscopes.

[0058] A "sound acquisition array" refers to an acquisition structure composed of multiple microphone units arranged in a specific geometric layout, used to simultaneously acquire sound pressure signals from multiple spatial points in a sound field to achieve sound source localization.

[0059] "Miniature microphone unit" refers to a tiny acoustic-to-electrical conversion device, which may include electret microphones or MEMS microphones, used to convert sound wave signals into electrical signals.

[0060] The "front-end analog conditioning module" refers to the circuit module connected to the microphone output, which is used to amplify, filter, and control the gain of audio signals to improve the signal-to-noise ratio.

[0061] The "analog-to-digital converter module" is used to convert analog audio signals into digital signals.

[0062] The "clock synchronization module" is used to ensure the time consistency of multi-channel signal sampling and to guarantee the phase correspondence of multi-microphone data.

[0063] "Shielding and vibration damping structure" refers to mechanical and electromagnetic protection structures used to prevent environmental electromagnetic interference and mechanical vibration from affecting the accuracy of sound acquisition, including metal shielding layers, rubber vibration damping pads, etc.

[0064] The “data interface module” refers to the hardware interface used to realize the digital signal transmission between the sound acquisition array and the terminal, which can be USB, Ethernet or other high-speed communication interfaces.

[0065] A “processor” is a computing unit used to perform computing, analysis and data processing tasks, and may include CPU, GPU, DSP or AI-specific chips.

[0066] "Memory" refers to a storage device used to store program instructions, model parameters and test data, which may include ROM, RAM or non-volatile memory.

[0067] The "noise fault database" refers to a database that stores known circuit board noise samples and their corresponding electrical signal characteristics. It is used to compare the samples with the signal under test in order to identify the type and cause of the noise.

[0068] "AI algorithm" or "AI model" refers to an algorithmic program that uses artificial intelligence technology to achieve signal recognition, feature extraction and pattern matching, which may include, but is not limited to, machine learning models, neural network models, pattern recognition models or combinations thereof.

[0069] Reference Figure 1 In one embodiment of the present invention, the noise detection method for a circuit board includes steps S100-S400, wherein: In step S100, in this embodiment, the test system or terminal sends a power-on command signal to the circuit board under test to control the power module to be turned on, so that the circuit board is in working state.

[0070] When powered on, the electronic components on the circuit board (such as inductors, capacitors, crystal oscillators, power modules, etc.) begin to operate, which may generate howling or abnormal noise caused by electromagnetic coupling, mechanical vibration or circuit resonance.

[0071] Step S200: After the circuit board is powered on and running stably, the sound generated at different locations on the circuit board is synchronously acquired through a sound acquisition array set around the circuit board via multiple channels.

[0072] The sound acquisition array includes multiple microphone units distributed according to a preset positional relationship. Each microphone unit acquires sound wave signals at the same sampling rate, and after front-end amplification and filtering, a multi-channel audio signal is formed.

[0073] The audio signal contains the sound pressure characteristics of the circuit board at different spatial locations, which can reflect the noise distribution characteristics generated by the circuit board during operation.

[0074] In step S300, the audio signal is transmitted to the terminal after analog-to-digital conversion. The terminal includes a processor and a memory, and the memory contains a pre-set artificial intelligence algorithm model for analyzing and recognizing the audio signal.

[0075] The processor first calculates the arrival time difference between each channel of the multi-channel audio signal to obtain the approximate area of ​​the noise source (i.e., the coarse localization area). Then, delay compensation and superposition are performed on audio signals from different directions within this area, and the precise location area of ​​the sound source is determined based on the power peak of the superimposed signal. Furthermore, the terminal can establish a mapping relationship between the precisely located area and the device layout diagram of the circuit board, and judge the area through a confidence verification model. When the confidence level is higher than a preset threshold, the area is identified as a noise area.

[0076] Therefore, through multi-layered analysis using AI models, it is possible to automatically and accurately locate the areas on a circuit board that generate noise.

[0077] Step S400: After determining the noise region, control the oscilloscope probe to connect to the circuit node corresponding to the noise region and collect the voltage and current signals at the node.

[0078] The terminal performs frequency domain, time domain, and statistical feature analysis on the collected signals, extracting multi-dimensional parameters including harmonic peak value, rising edge jitter, amplitude overshoot, and spectral flatness.

[0079] The above parameters are input into the artificial intelligence model for processing and compared with the preset noise fault waveform database. When the matching result exceeds the preset threshold, the corresponding node is identified as a noise source.

[0080] The terminal outputs the detection results, including the spatial location information of the noise source and the corresponding component information.

[0081] Furthermore, refer to Figure 2 This embodiment provides a method for locating noise regions on a circuit board based on an AI model, including: Step S301: When the circuit board is powered on, multi-channel audio signals are acquired by an audio acquisition array arranged around the circuit board.

[0082] The sound acquisition array consists of multiple microphone units arranged in a preset positional relationship. The signals of each channel are acquired at the same sampling frequency to ensure time synchronization.

[0083] The acquired raw audio signal is amplified, filtered, and converted from analog to digital by the front-end conditioning circuit before being input into the AI ​​model in the terminal.

[0084] The AI ​​model can be a sound source localization model based on machine learning or deep learning, which is used to estimate the time difference, analyze the sound energy distribution, and infer the spatial location of the input audio data.

[0085] In step S302, the processor in the terminal performs cross-correlation analysis on the audio signals of each channel and extracts the time difference of arrival (TDOA) between different channels.

[0086] The time difference between each pair of microphones reflects the spatial geometric relationship between the sound source and the microphone array. Based on the TDOA data of multiple channel pairs, the terminal calculates the preliminary position distribution of the sound source on the circuit board.

[0087] By intersecting multiple sets of TDOA results, the approximate location range of the sound source can be obtained, i.e., the "coarse localization area".

[0088] In practical applications, the coarse localization area is usually a rectangular or elliptical area with an error in the center coordinates within the centimeter range. This can effectively limit the approximate range of the sound source and reduce the amount of calculation required for subsequent analysis.

[0089] In step S303, after the coarse positioning area is determined, the terminal scans multiple possible directions based on the array geometric parameters.

[0090] For each scanning direction, the system calculates the theoretical propagation time difference of each channel signal in that direction and performs corresponding delay compensation on the audio signal to align the signals from the target direction on the time axis.

[0091] Then, the signals from each channel after delay compensation are superimposed, and the power or energy distribution of the superimposed signal is calculated.

[0092] When the superimposed signal power reaches its peak, the corresponding direction is the location where the sound source is most likely to exist.

[0093] The system defines the area around the coordinate point where the power peak is located as the "precise positioning area".

[0094] Furthermore, in one embodiment, after completing the analysis of the audio signal and obtaining the precise location area of ​​the sound source, the system further performs the following steps: The terminal stores the layout information of the circuit board to be tested, which includes the two-dimensional coordinate system of the circuit board, the spatial position of each component, the component type and silkscreen number.

[0095] Step S304: After the AI ​​model outputs the spatial coordinates of the precisely located area, the terminal converts the acoustic coordinate system into a board surface coordinate system corresponding to the circuit board layout diagram based on the geometric calibration parameters of the acoustic acquisition array.

[0096] This coordinate transformation allows for the precise determination of the circuit area and related components corresponding to the location on the layout diagram.

[0097] In a preferred embodiment, the mapping relationship can be established through spatial affine transformation or perspective projection transformation, thereby ensuring the consistency between acoustic coordinates and actual physical coordinates.

[0098] After the mapping is completed, the AI ​​verification module in the terminal calculates the reliability of the mapping result based on the historical test samples of the circuit board and the feature data of the components, and obtains the confidence value.

[0099] This confidence level is used to characterize the degree of consistency between the current acoustic localization results and the actual structure of the circuit board.

[0100] Optionally, the AI ​​model can calculate the confidence level by considering the following factors: (1) The ratio of the peak power of the sound source localization signal to the average power of the surrounding area; (2) The ratio of the overlapping area between the positioning area and the known component boundary; (3) The probability of noise occurrence of this type of device in the circuit board during historical testing; (4) Temporal stability of multi-frame acoustic detection results.

[0101] The confidence level obtained after comprehensive weighting ranges from 0 to 1, with a larger value indicating a more reliable test result.

[0102] Step S305: When the calculated confidence level is greater than a preset first threshold (e.g., 0.85), the system determines the precise positioning area as a noise area; If the confidence level is lower than the threshold, the current detection result is deemed unreliable, and acoustic acquisition can be repeated or the detection range expanded.

[0103] By following the steps above, misjudgments caused by environmental noise or non-device vibration can be effectively eliminated, thereby improving the accuracy and stability of noise localization results.

[0104] Optionally, in one embodiment, after determining the noise region through acoustic detection, the system further performs an electrical signal detection step to achieve accurate identification of the noise source, which includes the following exemplary processes: The probe unit of the terminal control oscilloscope is connected to a circuit node in the noisy area, such as the crystal oscillator output terminal, the inductor input terminal, the two ends of the capacitor, or the power amplifier output terminal.

[0105] The oscilloscope samples the voltage and current signals of the node in real time, preferably at a sampling frequency of no less than 1GSa / s, in order to capture high-frequency howling and harmonic signals.

[0106] The collected data is filtered by the oscilloscope's internal signal processing module to remove DC components and low-frequency interference, resulting in a waveform signal that reflects noise characteristics.

[0107] The oscilloscope or terminal performs multidimensional feature extraction on the acquired electrical signals.

[0108] Specifically, the system extracts signal features from three dimensions: Frequency domain characteristics: including parameters such as harmonic peak value, spectral flatness, and dominant frequency offset, used to reflect the abnormal resonance of the circuit in a specific frequency band; Time-domain characteristics: including parameters such as rising edge jitter, amplitude overshoot, periodic pulse characteristics, and noise fluctuation period, used to characterize the stability of the signal on the time axis; Statistical characteristics include the signal's mean, standard deviation, peak-to-peak rate of change, etc., which are used to describe the overall fluctuation trend of the signal.

[0109] The aforementioned characteristic parameters constitute a multi-dimensional parameter set, which can comprehensively reflect the dynamic behavior and potential abnormal characteristics of electrical signals.

[0110] The terminal is pre-loaded with a noise fault waveform database, which stores feature templates for various typical component noise faults, including but not limited to: High-frequency whistling caused by crystal oscillator load capacitance mismatch; Periodic noise caused by inductor magnetic saturation; Spikes caused by power amplifier self-oscillation; Abnormal amplitude fluctuations caused by capacitor aging, etc.

[0111] The terminal inputs the currently measured multi-dimensional parameter set into an artificial intelligence model (such as a pattern recognition or clustering model) for feature comparison and calculates its similarity with the database template.

[0112] When the similarity result calculated by the AI ​​model is higher than the preset second threshold (e.g., 0.8), the system determines that the current circuit node is consistent with a certain typical noise pattern in the database, thereby identifying the node as a noise source.

[0113] If the similarity is below the threshold, the terminal can continue to detect adjacent nodes or resample to further confirm the source of the noise.

[0114] To further explain, the process of performing feature analysis on electrical signals and forming a multi-dimensional parameter set also includes: After the oscilloscope completes the acquisition of node signals, the terminal performs Fast Fourier Transform (FFT) or other spectral analysis on the acquired electrical signals to obtain the frequency distribution characteristics of the signals.

[0115] The terminal extracts the following feature parameters from the spectrum data: Harmonic peak value: refers to the amplitude of each harmonic component other than the fundamental frequency, used to characterize the resonant state of a circuit. When the harmonic peak value increases abnormally, it usually indicates the presence of device squealing or load mismatch. Spectral flatness refers to the uniformity of signal spectral energy distribution across frequency bands, reflecting whether there is concentration or abrupt change in the signal spectrum. A decrease in spectral flatness indicates that the energy in a certain frequency band is too high, which may correspond to increased local noise.

[0116] In some embodiments, the system can also extract additional parameters such as the main frequency drift and the frequency band energy ratio to further improve the expressive power of the feature dimensions.

[0117] The terminal performs time-domain analysis on the original waveform of the electrical signal and calculates the following typical parameters: Rising edge jitter: refers to the degree of fluctuation in the timing of the rising edge of a signal, used to reflect timing stability. If the rising edge jitter exceeds a preset threshold, it may indicate that the circuit has self-oscillation or power supply noise coupling; Amplitude overshoot: refers to the proportion of a signal that overshoots the rated amplitude during the rise or fall of the signal, and is used to judge the impedance matching of the circuit. Periodic noise pulses: refer to periodic spikes or glitches that appear in a waveform. This characteristic is often caused by oscillator whistling or inductor resonance.

[0118] By extracting the above parameters, the fluctuation characteristics of the signal in the time domain can be quantitatively described.

[0119] The terminal performs statistical calculations on the time-domain signal sequence to obtain the following statistical characteristic parameters: Standard deviation: reflects the dispersion of signal fluctuation amplitude. The larger the standard deviation, the worse the signal stability. Peak variation: refers to the difference between the maximum and minimum values ​​of a signal within a sampling period, reflecting the fluctuation range of instantaneous voltage or current.

[0120] In addition, the system can calculate additional statistics such as root mean square value, skewness, and kurtosis to enhance the feature discrimination capability.

[0121] The terminal combines the various feature parameters extracted from the frequency domain, time domain, and statistical domain in a fixed order or by index encoding to form a multi-dimensional parameter set.

[0122] This set can be represented in vector form: ; in, It is a multi-dimensional parameter set. Represents frequency domain characteristic parameters, Represents time-domain characteristic parameters, Represents statistical characteristic parameters; Indicates the parameter index.

[0123] The terminal performs standardization and normalization on the parameter set to eliminate amplitude differences caused by different test conditions, so that the set can be directly used for feature comparison and similarity calculation of subsequent AI models.

[0124] Optionally, in another embodiment, an infrared scanning step is added before performing waveform detection to identify and analyze the temperature distribution in the noise area of ​​the circuit board, so as to further improve the accuracy and reliability of noise source localization.

[0125] After obtaining the noise area through the audio acquisition array, the terminal controls the thermal imager to perform infrared scanning on the noise area.

[0126] The thermal imager includes an infrared sensing unit and an imaging processing unit, used to collect infrared radiation information on the surface of the circuit board and generate a temperature distribution map.

[0127] Temperature distribution maps are two-dimensional or three-dimensional grayscale or pseudo-color images that reflect the surface temperature distribution at different locations on the circuit board.

[0128] The terminal analyzes the temperature distribution map using image processing algorithms and identifies points or areas with temperatures significantly higher than the surrounding areas as "abnormal heat points".

[0129] In a preferred embodiment, the identification of abnormal hot spots can be achieved by threshold segmentation, temperature gradient analysis, or edge detection algorithm.

[0130] The terminal performs spatial alignment and fusion analysis on the abnormal heat point information obtained by thermal imaging and the noise area located by the audio acquisition array.

[0131] Through coordinate calibration, the system can unify the coordinates of thermal imaging images and the acoustic positioning coordinates of the circuit board into the same planar coordinate system.

[0132] The terminal calculates the distance between the abnormal heat source and the center point of the noise area, and determines the spatial relationship between the two.

[0133] If no obvious abnormal heat point is detected during the infrared scanning process, or if the temperature difference of the detection result is less than the preset temperature difference threshold (e.g., 0.5℃), the system will assume that the current noise source is not caused by device overheating or abnormal power dissipation, but may be electromagnetic oscillation or mechanical coupling type howling.

[0134] At this point, the terminal automatically reduces the weight of thermal imaging results in multi-source fusion, retaining only acoustic detection and oscilloscope waveform detection results as the main judgment criteria.

[0135] This dynamic weight adjustment mechanism can automatically optimize decision-making strategies under different types of noise scenarios, preventing thermal imaging from misjudging non-thermal source noise.

[0136] When the abnormal heat source and the noise area coincide in space or the distance is less than the preset third threshold (e.g., 5mm), the system considers the sound source to be highly correlated with the thermal anomaly.

[0137] At this point, the terminal automatically increases the confidence weight of the noise source determination and prioritizes outputting this area as the main noise source.

[0138] Based on the fusion analysis results, the terminal outputs a detection report that includes noise source location information, heating status, and overall confidence level.

[0139] The report can indicate the temperature of abnormal heating points, the coordinates of the sound source location, and the confidence level, which facilitates subsequent fault analysis and maintenance operations.

[0140] Optionally, refer to Figure 3 Another embodiment of the present invention provides a signal detection component for performing a detection method. The signal detection component of this embodiment includes: Sound acquisition array, oscilloscope, thermal imager, and terminal; The sound acquisition array, oscilloscope, and thermal imager establish signal transmission relationships with the terminal to transmit the acquired multi-source data to the terminal for unified processing.

[0141] The various components work together to form a multi-channel signal acquisition and fusion analysis system, enabling synchronous detection and intelligent analysis of acoustic, electrical, and thermal information.

[0142] The sound acquisition array is used to acquire acoustic signals generated by the circuit board when it is powered on.

[0143] The sound acquisition array includes multiple miniature microphone units distributed according to a preset array position relationship, preferably MEMS microphones.

[0144] The position of each microphone unit relative to the array reference point is calibrated before installation to ensure spatial consistency of the acquired signals.

[0145] The analog audio signal output from the microphone unit is amplified and filtered at the front end before being input to the analog-to-digital converter module to achieve multi-channel synchronous sampling.

[0146] The audio acquisition array is connected to the terminal via a high-speed data interface (such as USB or Ethernet) to ensure real-time transmission of audio data.

[0147] During the test, the sound acquisition array was placed at a certain distance above the circuit board (e.g., 5cm to 10cm) to create a stable sound field acquisition environment.

[0148] The array can capture sound pressure changes in different areas of the circuit board, enabling multi-channel sound source localization.

[0149] An oscilloscope is used to acquire electrical signals and analyze waveforms at circuit nodes in noisy areas.

[0150] It includes: Probe unit: can be connected to a designated node on the circuit board to collect voltage and current signals; Signal processing unit: performs digital processing on the acquired signals, including filtering, sampling, spectrum analysis and timing analysis.

[0151] The oscilloscope is connected to the terminal via a data bus or LAN network, transmitting sampled data to the terminal's AI processing module in real time.

[0152] The terminal extracts multi-dimensional feature parameters based on the acquired waveform signals and compares them with the built-in noise fault waveform database. It then uses similarity calculation to determine whether the current node is a noise source.

[0153] This module can effectively identify howling noise caused by electrical problems such as component parameter drift, inductor oscillation, or crystal mismatch.

[0154] Thermal imagers are used to perform infrared scanning of noisy areas on circuit boards to detect possible abnormal heating phenomena.

[0155] It includes: Infrared sensing unit: used to collect infrared radiation information from the surface of the circuit board and convert it into a temperature signal; Imaging processing unit: performs interpolation, noise reduction, and pseudo-color processing on the temperature signal to generate a temperature distribution map.

[0156] The thermal imager can connect to the terminal via wired or wireless means to transmit infrared image data in real time.

[0157] The terminal analyzes the temperature distribution map using image algorithms and identifies areas with temperatures significantly higher than the average as abnormal heat points.

[0158] When the abnormal heating point coincides with or is close to the noise area obtained by acoustic localization in space, the system increases the confidence level of the noise source determination. If the thermal imaging results do not show obvious hot spots, their weight in the fusion analysis will be automatically reduced to prevent misjudgment.

[0159] The terminal is the core control and analysis unit of the signal detection component.

[0160] It includes a processor and memory. The processor can be a CPU, GPU, or AI-specific acceleration chip, and the memory stores program code for noise detection, AI algorithm models, and a noise fault database.

[0161] The terminal maintains real-time data exchange with the sound acquisition array, oscilloscope, and thermal imager, and performs the following operations: It receives multi-channel audio signals from the sound acquisition array and analyzes the signals using an AI model to calculate the coarse and precise localization areas of the sound source. Receive electrical signal waveform data acquired by an oscilloscope, perform multi-dimensional feature extraction and pattern matching on it, and determine the noise source nodes; The system receives temperature distribution maps acquired by a thermal imager and performs coordinate mapping and fusion analysis on abnormal heating points and sound source areas. Based on the fusion of multi-source data, the comprehensive confidence level is calculated, and the noise source location information and judgment report are output.

[0162] The terminal can display the analysis results in a graphical interface, including circuit board layout diagrams, sound pressure distribution diagrams, temperature distribution diagrams, and waveform characteristic curves, which facilitates technicians in locating and correcting errors.

[0163] Furthermore, the sound acquisition array includes multiple miniature microphone units, which are distributed on the array base according to a preset array position relationship.

[0164] In a preferred embodiment, the number of microphone units is 8 to 16, arranged in a matrix, circular or linear array, and the center-to-center distance between the microphones is in the range of 1 cm to 2 cm.

[0165] The array base is equipped with a reference point. The three-dimensional spatial position of each microphone unit relative to the reference point is calibrated with high precision at the factory and a position coordinate model is established in the terminal.

[0166] With this calibration information, the system can calculate the sound source location based on the arrival time difference of signals from different channels, achieving sub-centimeter spatial resolution.

[0167] The microphone unit is preferably a MEMS miniature microphone, which features high sensitivity, low noise, and anti-interference performance. Each unit has an independent output port to ensure signal integrity and channel independence.

[0168] This array structure enables it to accurately capture local acoustic field characteristics of a circuit board in noisy environments, making it suitable for noise detection scenarios with high-density component layouts.

[0169] The front-end analog conditioning module is connected to the output of each microphone unit and is used to amplify, filter and adjust the gain of the original audio signal with low noise.

[0170] Specifically: Low-noise amplifiers (LNAs) are used to boost signal levels while suppressing thermal noise and power ripple. The filtering circuit adopts a bandpass filter structure to filter out invalid signals below 20Hz and above 20kHz, so as to preserve the noise frequency band of the circuit board; Programmable gain amplifiers (PGAs) are used to automatically adjust the signal amplitude according to different test environments to ensure stable input dynamic range.

[0171] This module can effectively improve the signal-to-noise ratio, providing high-quality signal input for subsequent analog-to-digital conversion and AI analysis.

[0172] To ensure time consistency of data across all channels, the array is equipped with analog-to-digital conversion and clock synchronization modules.

[0173] This module includes a multi-channel ADC chip and a master clock source, used for synchronous sampling of conditioned analog audio signals.

[0174] Analog-to-digital converter (ADC): The sampling rate per channel is preferably no less than 48kHz, and the quantization accuracy is 16 bits or higher to ensure that high-frequency howling characteristics are captured.

[0175] Clock synchronization circuit: A unified reference clock or distributed phase-locked loop synchronization mechanism is adopted to ensure that the sampling error between channels is less than 1 / 10 of the sampling period, thereby guaranteeing the accuracy of the time difference of arrival (TDOA) calculation.

[0176] Timestamp generation unit: Adds a timestamp to each frame of data for subsequent terminal to perform sound source localization and signal correlation analysis.

[0177] This module enables the sound acquisition array to achieve multi-channel synchronous sampling at the millisecond or even microsecond level, providing a precise temporal information foundation for AI algorithms.

[0178] To reduce the impact of external interference on detection accuracy, the sound acquisition array is surrounded by a shielding and vibration damping structure, including: Electromagnetic shielding layer: A conductive metal mesh or copper-plated shell is used to cover the periphery of the array to suppress electromagnetic radiation from the circuit board and environmental radio frequency interference; Mechanical vibration damping layer: Rubber, silicone or foam damping material is placed between the array base and the shell to absorb external vibration and impact and prevent mechanical noise from mixing into the acquired signal; Airflow field isolation cover: A transparent soundproof cover can be installed when needed to reduce airflow noise and acoustic crosstalk.

[0179] With the above structure, the array can maintain stable signal acquisition performance in high-interference experimental environments.

[0180] The data interface module is used to achieve high-speed output of multi-channel digital signals and communication with the terminal.

[0181] This module includes, but is not limited to, one of the following connection methods: USB 3.0 interface for high-speed data transmission in small-scale channel arrays; Gigabit Ethernet interface for medium to large-scale acquisition arrays; The PCIe interface is used in laboratory-level real-time acquisition systems, with bandwidth up to several GB / s.

[0182] The data interface module is also responsible for packaging, synchronously marking, and transmitting multi-channel data to the terminal, enabling the terminal to receive audio streams in real time and perform sound source localization calculations.

[0183] Furthermore, the processor of the terminal is configured to perform the following steps: The terminal receives multi-channel audio signals from the sound acquisition array via a high-speed data interface.

[0184] The system first performs Time Difference of Arrival (TDOA) estimation to calculate the arrival time difference of the audio signals in each channel.

[0185] By combining the array's geometric position parameters, the coarse localization region of the sound source is solved using the least squares method or backpropagation algorithm. This region is generally a local area in a two-dimensional plane or a cone-shaped area in three-dimensional space.

[0186] After obtaining the coarse positioning area, the terminal performs delay compensation and superposition processing on the audio signals from different directions.

[0187] Specifically, the system uses hypothetical sound sources at different directional angles as centers to perform phase correction and time delay compensation on the signals of each channel. Then calculate the curve of the superimposed power of the compensated signal; the direction corresponding to the power peak is the precise positioning area.

[0188] This process is equivalent to a dynamic beamforming operation, which enhances the real sound source signal and suppresses background noise interference by focusing energy.

[0189] In a preferred embodiment, the terminal can utilize multi-threaded parallel computing to achieve real-time processing, ensuring detection response speed.

[0190] The terminal calls the AI ​​model to verify the confidence level of the aforementioned precisely located area.

[0191] The AI ​​model uses deep learning algorithms to comprehensively analyze the sound pressure distribution, spectral characteristics, and component distribution information in the circuit layout diagram of the signal, and calculates the probability that each location is a noise source.

[0192] The system filters candidate sound sources based on a confidence threshold. When the confidence level is greater than a preset first threshold (e.g., 0.8), the device at that location is determined to be a noise source.

[0193] The terminal further receives detection data from the oscilloscope and thermal imager: The oscilloscope provides voltage and current waveforms and spectrum analysis results; The thermal imager provides a temperature distribution map and coordinates of abnormal heating points.

[0194] The terminal integrates and analyzes acoustic positioning results, electrical signal characteristics, and temperature anomaly information.

[0195] The specific method is as follows: By mapping spatial coordinates, the sound source location point is compared with the waveform anomaly node and the temperature rise area, and the degree of overlap of the three is calculated.

[0196] If all three are concentrated in the same area, the confidence level increases significantly. If thermal imaging does not detect any abnormal heating, the weight of that channel in the fusion analysis is reduced to prevent misjudgment.

[0197] The terminal can output a test report on the display interface, which includes: Schematic diagram of the circuit board and markings of the sound source location points; Corresponding waveform analysis curves and spectral characteristics; Infrared temperature distribution map and marking of abnormal heating points; Explanation of overall confidence level and fault category.

[0198] Users can perform targeted repairs based on the test results, or export the test data to create a record document.

[0199] Optionally, in a preferred embodiment of the present invention, the noise detection method and signal detection component for circuit boards are applied to the quality inspection process of a PCB board in a communication device. During mass production testing, approximately 15% of the PCB boards exhibited a 2.4MHz high-frequency howling phenomenon. Manual inspection could not accurately pinpoint the location of the howling components, resulting in a rework rate as high as 18%, severely impacting production yield and inspection efficiency. To solve this problem, the comprehensive inspection method provided by the present invention, which combines acoustic detection, oscilloscope analysis, and thermal imaging, was used to automatically locate and verify the problematic samples.

[0200] In actual testing, a miniature microphone array with eight channels was fixed approximately 5 cm directly above the PCB board, ensuring a pre-defined spatial layout of the array channels relative to the board. Upon power-up, the circuit board captured a howling signal with a center frequency of 2.401 MHz in just about 0.3 seconds. Based on the acquired multi-channel audio signals, the terminal system used the Generalized Cross-Correlation Band Phase Transform (GCC-PHAT) algorithm to calculate the Time Difference of Arrival (TDOA), thus obtaining a coarse localization area for the noise source. The calculation results showed that the source of the howling signal was roughly located within a range of 12.1 to 12.5 cm in the X direction and 8.3 to 8.7 cm in the Y direction on the circuit board, corresponding to the silkscreen area "U12" on the PCB, and its component was a crystal oscillator of model X1-2400.

[0201] Based on this, the system further performs dynamic focused beamforming scanning within the coarse positioning area. By delay compensation and superposition of acoustic signals from different directions, the coordinates of the power peak point are obtained. Analysis results show that the precise location corresponding to the maximum sound pressure level is (X: 12.3cm, Y: 8.5cm). The pre-installed artificial intelligence model in the terminal verifies the confidence level of this location. After considering factors such as sound pressure distribution characteristics and circuit board layout information, the model outputs a confidence level of 97.2%, thus confirming that the U12 crystal oscillator is the noise source.

[0202] To further verify the acoustic test results, an oscilloscope was used to acquire waveforms at the output terminal (pin 2) of the U12 crystal oscillator. The oscilloscope sampling rate was set to 10 GSa / s. After analysis by an AI algorithm, the acquired voltage signal showed harmonic peaks of 4.802 MHz and 7.203 MHz in the frequency domain. In terms of time domain characteristics, the signal rise time jitter reached 62 picoseconds, significantly exceeding the normal range (≤30 picoseconds). The system input these characteristic parameters into a preset noise fault waveform database and performed similarity matching with typical templates. The database ultimately found a 94% match with the "crystal oscillator load capacitance mismatch" pattern, thus determining the fault type as crystal oscillator load capacitance parameter mismatch.

[0203] Subsequently, an infrared thermal imager was used for auxiliary detection of the U12 area. The thermal imager had a resolution of 640×480 and a temperature measurement accuracy of ±0.1℃. The detection results showed that the temperature rise in the area surrounding U12 was only about 1.2℃, and no significant heating phenomenon was observed. During the fusion analysis phase, the system automatically reduced the weight of the thermal imaging results in the comprehensive judgment, and used acoustic and electrical signal characteristics as the main basis for fusion calculation. The comprehensive score was higher than the system threshold, and the U12 crystal oscillator was finally confirmed as the noise source causing the 2.4MHz howling.

[0204] Based on the test results, targeted modifications were made to the circuit board. By comparing the crystal oscillator load parameters, it was confirmed that the original capacitor C36 was a 10pF model, which was incompatible with the resonant conditions of the crystal oscillator. After replacing it with a 20pF capacitor, the circuit board's whistling phenomenon completely disappeared during retesting. The 2.4MHz signal was no longer captured in the acoustic test, and the corresponding harmonic peaks in the oscilloscope waveform also disappeared. The rise time jitter was reduced to 22 picoseconds, returning to the normal operating range.

[0205] Through the above-described embodiments, it has been verified that the noise detection method and signal detection component of the present invention can achieve rapid and accurate localization of high-frequency howling sources in complex PCB electromagnetic acoustic environments. The entire detection process is highly automated, with a response time of less than 1 second and a localization error of less than 2 centimeters, significantly improving the detection efficiency and diagnostic accuracy of circuit boards, effectively reducing human error and rework costs, and verifying the practical application value and reliability of this technical solution in the field of communication equipment production and testing.

[0206] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A noise detection method for a circuit board, characterized in that, The method includes: Send a power-on signal to put the circuit board into a powered state; After entering the power-on state, sound is collected from the circuit board in the power-on state to obtain an audio signal; The audio signal is analyzed using an AI model to identify the noise region; Waveform detection is performed on the circuit nodes in the noise region to locate the noise source, and the detection result including the location information of the noise source is output.

2. The noise detection method for a circuit board as described in claim 1, characterized in that, The audio signal is a multi-channel audio signal.

3. The noise detection method for a circuit board as described in claim 2, characterized in that, The analysis of the audio signal based on the AI ​​model yields the following noise regions: The audio signal is input into the AI ​​model; The arrival time difference between each channel of the audio signal is calculated, and the coarse localization region of the noise source is obtained based on the arrival time difference; Within the coarse positioning area, delay compensation and superposition are performed on audio signals from different directions to obtain a precise positioning area corresponding to the power peak.

4. The noise detection method for a circuit board as described in claim 3, characterized in that, The audio signal was analyzed using an AI model, and the resulting noise region also included: A mapping relationship is established between the precise positioning area and the circuit board layout diagram, and the confidence level of the mapping relationship is verified to obtain the confidence level. When the confidence level is greater than a preset first threshold, the precise positioning area is determined to be a noise area.

5. The noise detection method for a circuit board as described in claim 1, characterized in that, The step of using an oscilloscope to perform waveform detection on the circuit nodes in the noise region to locate the noise source includes: In the circuit nodes corresponding to the noise region, electrical signals are collected, and the signal characteristics of the electrical signals are analyzed to form a multi-dimensional parameter set. The multi-dimensional parameter set is input into the AI ​​model for processing and matched with a preset noise fault waveform database. When the matching result is higher than a preset second threshold, the corresponding circuit node is identified as a noise source.

6. The noise detection method for a circuit board as described in claim 5, characterized in that, The step of performing feature analysis on the signal characteristics of the electrical signal to form a multi-dimensional parameter set includes: Extracting frequency domain features, time domain features, and statistical features of electrical signals; The characteristic parameters of the frequency domain features include at least harmonic peak value and spectral flatness; the characteristic parameters of the time domain features include at least rising edge jitter, amplitude overshoot and periodic noise pulse; and the characteristic parameters of the statistical features include at least standard deviation and peak value variation. All feature parameters are constructed into a multi-dimensional parameter set.

7. The noise detection method for a circuit board as described in claim 1, characterized in that, The method further includes: Before performing the waveform detection, the noise area is scanned with an infrared sensor to obtain the temperature distribution map of the area and identify abnormal hot spots. The abnormal hot spot is used as auxiliary information and fused with the noise area obtained by the audio acquisition array for analysis. When the abnormal hot spot coincides with the noise area or is close to the preset third threshold, the confidence of the noise source determination is improved. If the abnormal heat points do not appear significantly, the weight of the thermal imaging results is reduced.

8. A signal detection component, performing the noise detection method for a circuit board as described in any one of claims 1 to 7, characterized in that, The signal detection component includes a sound acquisition array, an oscilloscope, a thermal imager, and a terminal; The sound acquisition array, oscilloscope, and thermal imager all establish a signal transmission relationship with the terminal; The sound acquisition array includes multiple miniature microphone units distributed according to a preset array position relationship, used to acquire audio signals from different positions on the circuit board; The terminal includes a processor and a memory, the processor being configured to use an AI model and an AI model to determine the noise region of the sound source on the circuit board; The oscilloscope includes a probe unit and a signal processing unit. The probe unit is used to connect to the circuit node corresponding to the noise area. The signal processing unit is used to perform spectrum and timing analysis on the acquired voltage and current signals, and match them with a preset fault waveform database based on an artificial intelligence model to identify the noise source. The thermal imager includes an infrared sensing unit and an imaging processing unit, which are used to perform infrared scanning on the noise area, generate a temperature distribution map, and fuse abnormal hot spots as auxiliary information with the acoustic localization results to improve the confidence of noise source identification.

9. The signal detection component as described in claim 7, characterized in that, The sound acquisition array includes: Multiple miniature microphone units are distributed according to a preset array position relationship. The spatial position of each microphone unit relative to the array reference point is pre-calibrated to synchronously collect audio signals from different positions when the circuit board is powered on. The front-end analog conditioning module is connected to the output of each microphone unit and is used to amplify, filter, and adjust the gain of the audio signal with low noise. The analog-to-digital conversion and clock synchronization modules are connected to the front-end analog conditioning module and are used to synchronously sample the audio signals of each channel and output digital signals with timestamps. A shielding and vibration damping structure is installed around the sound acquisition array to suppress environmental mechanical vibration and electromagnetic interference; The data interface module is used to transmit the multi-channel digital signals to the terminal in the form of a high-speed data bus in order to perform sound source localization.

10. The signal detection component as described in claim 7, characterized in that, The terminal includes: A processor and a memory, wherein the memory is used to store artificial intelligence algorithm programs and a preset noise fault database; The processor is configured to: The system receives audio signals from a sound acquisition array, performs time difference estimation on the audio signals, and obtains a coarse localization area of ​​the noise source. Based on the coarse positioning area, delay compensation and superposition are performed on audio signals from different directions to obtain the precise positioning area corresponding to the power peak. The AI ​​model is invoked to verify the confidence level of the precisely located area, and the noise source device is determined by combining the circuit board layout diagram; The system receives detection data from an oscilloscope and a thermal imager, performs fusion analysis on the detection data and acoustic detection results, and outputs detection results including noise source location information based on the fusion results.

Images