A PCBA testing method and system

Abstract

This invention relates to a PCBA testing method and system. The invention includes: injecting AC excitation signals into preset test points on the PCBA when it is not powered on; acquiring a first voltage-current response curve; constructing a corresponding feature vector and inputting the feature vector into a pre-trained machine learning classification model to obtain a power-down diagnostic result; acquiring first electrical signals of key nodes when the PCBA is powered on; constructing a first feature vector to characterize the operating state of each key node; inputting the first feature vector into a pre-trained isolated forest model to obtain a power-on diagnostic result; establishing a test performance database recording test item results and fault characteristics based on the power-down diagnostic result and the power-on diagnostic result; dynamically reordering unexecuted test items and updating the current test sequence based on the historical fault frequency and real-time acquired fault characteristics in the test performance database. This achieves intelligent testing and improves testing efficiency.

Description

Technical Field

[0001] This invention relates to the field of electronic manufacturing inspection technology, and in particular to a PCBA inspection method and system. Background Technology

[0002] Currently, as electronic products continue to evolve towards higher density, greater complexity, and miniaturization, manufacturing defects in printed circuit board assemblies (PCBAs) are becoming increasingly diverse and concealed. Traditional testing methods are no longer sufficient to meet the quality requirements of modern electronic manufacturing. Existing automated testing equipment generally employs static testing processes with a fixed sequence, failing to dynamically adjust test paths based on real-time fault characteristics, resulting in low overall testing efficiency. Furthermore, test data is isolated, making it difficult to support closed-loop optimization and quality traceability in the manufacturing process. Summary of the Invention

[0003] Therefore, the present invention provides a PCBA testing method and system that realizes intelligent testing process and improves testing efficiency.

[0004] To solve the above technical problems, the present invention provides a PCBA inspection method, comprising:

[0005] When the PCBA is not powered on, inject an AC excitation signal into the preset test points on the PCBA;

[0006] Based on the AC excitation signal, the first voltage-current response curves corresponding to each test point are acquired in real time.

[0007] Each first voltage-current response curve is compared with its corresponding second voltage-current response curve. Based on the comparison results, feature parameters characterizing electrical characteristics are extracted from each first voltage-current response curve to construct a corresponding feature vector. The feature vector is then input into a pre-trained machine learning classification model to identify faults including open circuit, short circuit, component parameter errors, and leakage, thereby obtaining a power-off diagnosis result. The second voltage-current response curve is a standard voltage-current response curve of the test point under normal conditions stored in a pre-established database.

[0008] When the PCBA is powered on, the first electrical signal of the key nodes is acquired; the key nodes include the power rail, clock line, and data bus.

[0009] Statistical features, time-domain features, and frequency-domain features are extracted from the first electrical signal to construct a first feature vector that characterizes the operating state of each key node.

[0010] The first feature vector is input into a pre-trained isolated forest model, anomaly score is calculated based on average path length, and normal or abnormal judgment and corresponding dynamic functional abnormality type are output according to the comparison result of anomaly score and preset threshold, so as to obtain power-on diagnosis result.

[0011] Based on the power-down diagnostic results and the power-on diagnostic results, a test performance database is established that records test item results and fault characteristics. Based on the historical fault frequency and real-time acquired fault characteristics in the test performance database, the test items that have not yet been executed are dynamically reordered, and the current test sequence is updated.

[0012] In one embodiment of the present invention, the statistical characteristics include the signal's mean, variance, standard deviation, peak-to-peak value, skewness, and kurtosis;

[0013] The time-domain features include rise time, fall time, pulse width, and duty cycle;

[0014] The frequency domain features include the main frequency, harmonic component energy, and total spectral energy obtained based on the Fast Fourier Transform.

[0015] In one embodiment of the present invention, offline training is also included, comprising:

[0016] The second electrical signal of key nodes of multiple known normal PCBAs during power-on operation is collected. Based on the second electrical signal, the second feature vector is extracted and a feature database containing only normal data is constructed. An isolated forest model is trained based on the feature database and the trained model is stored locally or in the cloud.

[0017] In one embodiment of the present invention, based on the historical fault frequency and real-time acquired fault characteristics in the test performance database, the unexecuted test items are dynamically reordered, and the current test sequence is updated, including:

[0018] Before starting the test, the test items are sorted according to the historical failure frequency of each test item recorded in the test performance database to generate an initial test sequence;

[0019] When the current test item is executed and a fault is detected, obtain the fault characteristics corresponding to the current test item;

[0020] Based on the fault characteristics, find one or more test items associated with the fault characteristics in the fault characteristic association graph;

[0021] In the sequence of test items that have not yet been executed, the test item associated with the fault feature is moved to the beginning of the sequence to generate an updated test sequence.

[0022] In one embodiment of the present invention, in the test performance database, each test item is assigned a unique name, historical failure frequency, and execution result, and a fault feature association diagram is pre-configured so that each type of fault feature corresponds to one or more associated test items.

[0023] The present invention also provides a PCBA inspection system, comprising:

[0024] The power-down diagnostic unit is used to inject AC excitation signals into preset test points on the PCBA when the PCBA is not powered on.

[0025] Based on the AC excitation signal, the first voltage-current response curves corresponding to each test point are acquired in real time.

[0026] Each first voltage-current response curve is compared with its corresponding second voltage-current response curve. Based on the comparison results, feature parameters characterizing electrical characteristics are extracted from each first voltage-current response curve to construct a corresponding feature vector. The feature vector is then input into a pre-trained machine learning classification model to identify faults including open circuit, short circuit, component parameter errors, and leakage, thereby obtaining a power-off diagnosis result. The second voltage-current response curve is a standard voltage-current response curve of the test point under normal conditions stored in a pre-established database.

[0027] The power-on diagnostic unit is used to acquire the first electrical signal of key nodes when the PCBA is powered on and operating; the key nodes include the power rail, clock line, and data bus.

[0028] Statistical features, time-domain features, and frequency-domain features are extracted from the first electrical signal to construct a first feature vector that characterizes the operating state of each key node.

[0029] The first feature vector is input into a pre-trained isolated forest model, anomaly score is calculated based on average path length, and normal or abnormal judgment and corresponding dynamic functional abnormality type are output according to the comparison result of anomaly score and preset threshold, so as to obtain power-on diagnosis result.

[0030] The dynamic reordering unit is used to establish a test performance database that records test item results and fault characteristics based on the power-down diagnostic results and the power-on diagnostic results, and to dynamically reorder the test items that have not yet been executed based on the historical fault frequency and real-time acquired fault characteristics in the test performance database, and update the current test sequence.

[0031] The technical solution of the present invention has the following advantages compared with the prior art:

[0032] This invention discloses a PCBA testing method and system. When the PCBA is not powered on, it injects an AC excitation signal and collects the voltage-current response curves of test points to extract feature parameters. A machine learning classification model is then used to identify structural faults such as open circuits, short circuits, component parameter errors, and leakage, enabling precise location of static electrical anomalies in the power-off state. After the PCBA is powered on, electrical signals from key nodes such as the power rails, clock lines, and data buses are collected, and statistical, time-domain, and frequency-domain features are extracted. These features are then input into an isolated forest model for dynamic functional anomaly detection, identifying operational faults such as signal distortion and abnormal noise. The combination of these two methods constitutes a complete multimodal diagnostic system, achieving full coverage detection from structural electrical anomalies to dynamic functional anomalies, significantly improving the comprehensiveness and accuracy of fault identification.

[0033] This invention constructs a test performance database, recording the unique name, historical failure frequency, and execution results of each test item, and builds a fault feature association graph. During testing, when a fault is detected in a test item, the system immediately executes the test items related to that fault in advance based on the fault feature association graph, achieving real-time reordering of the test sequence. This enables the detection path to automatically adapt to the fault distribution of the tested sample, improves the speed of detecting critical faults, reduces invalid testing steps, shortens the overall testing cycle, and improves testing efficiency. Attached Figure Description

[0034] To make the content of this invention easier to understand, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings.

[0035] Figure 1 This is a flowchart of the PCBA testing method of the present invention.

[0036] Figure 2 This is a structural diagram of the PCBA testing system of the present invention. Detailed Implementation

[0037] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can better understand and implement the present invention. However, the embodiments described are not intended to limit the present invention.

[0038] In this invention, when directions (up, down, left, right, front, and back) are described, it is only for the convenience of describing the technical solution of this invention, and does not indicate or imply that the technical features referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0039] In this invention, "several" means one or more, "multiple" means two or more, "greater than," "less than," "exceeding," etc., are understood to exclude the stated number; "above," "below," "within," etc., are understood to include the stated number. In the description of this invention, the terms "first" and "second" are used only to distinguish technical features and should not be construed as indicating or implying relative importance, or implicitly indicating the number of indicated technical features, or implicitly indicating the order of the indicated technical features.

[0040] In this invention, unless otherwise explicitly defined, the terms "setting," "installing," and "connecting" should be interpreted broadly. For example, they can refer to a direct connection or an indirect connection through an intermediate medium; a fixed connection, a detachable connection, or an integrally formed connection; a mechanical connection, an electrical connection, or a connection capable of mutual communication; or the internal connection of two components or the interaction between two components. Those skilled in the art can reasonably determine the specific meaning of the above terms in this invention based on the specific content of the technical solution.

[0041] Example 1

[0042] Reference Figure 1 As shown, this embodiment provides a PCBA testing method, including:

[0043] S1. When the PCBA is not powered on, inject an AC excitation signal into the preset test points on the PCBA.

[0044] S2. Based on the AC excitation signal, the first voltage-current response curve corresponding to each test point is acquired in real time;

[0045] S3. Compare each of the first voltage-current response curves with the corresponding second voltage-current response curves. Based on the comparison results, extract feature parameters for characterizing electrical characteristics from each of the first voltage-current response curves, construct corresponding feature vectors, and input the feature vectors into a pre-trained machine learning classification model to identify faults including open circuit, short circuit, component parameter errors, and leakage current, and obtain power failure diagnosis results; wherein, the second voltage-current response curve is a standard voltage-current response curve of the test point under normal conditions stored in a pre-established database;

[0046] S4. When the PCBA is powered on, acquire the first electrical signal of the key nodes; the key nodes include the power rail, clock line, and data bus.

[0047] S5. Extract statistical features, time-domain features and frequency-domain features from the first electrical signal to construct a first feature vector to characterize the operating state of each key node;

[0048] S6. Input the first feature vector into the pre-trained isolated forest model, calculate the anomaly score based on the average path length, and output the normal or abnormal judgment and the corresponding dynamic functional abnormality type according to the comparison result of the anomaly score and the preset threshold to obtain the power-on diagnosis result.

[0049] S7. Based on the power-down diagnosis results and the power-on diagnosis results, establish a test performance database that records test item results and fault characteristics. Based on the historical fault frequency and real-time acquired fault characteristics in the test performance database, dynamically reorder the test items that have not yet been executed and update the current test sequence.

[0050] Specifically, the statistical characteristics include the signal's mean, variance, standard deviation, peak-to-peak value, skewness, and kurtosis;

[0051] The time-domain features include rise time, fall time, pulse width, and duty cycle;

[0052] The frequency domain features include the main frequency, harmonic component energy, and total spectral energy obtained based on the Fast Fourier Transform.

[0053] It should be noted that under normal conditions, the impedance, capacitance, inductance, and parasitic parameters corresponding to different test points all have a stable distribution. Therefore, after injecting an AC excitation signal of a specific frequency and amplitude, its voltage-current response curve exhibits repeatable amplitude changes, phase characteristics, and waveform morphology. After injecting an AC excitation signal with the same parameters into the PCBA under test, the system acquires the first voltage-current response curve of each test point in real time. Subsequently, this curve is aligned with the standard curve point by point or within a characteristic range to compare the differences, mainly in the following directions:

[0054] Amplitude deviation analysis determines whether the voltage and current amplitudes deviate significantly from the standard curve, and is used to reflect resistance changes and leakage trends.

[0055] Phase difference analysis identifies capacitor and inductor bias, welding defects, and parasitic parameter anomalies by comparing phase angle changes;

[0056] Curve shape analysis identifies poor contact, partial short circuit, and component failure by detecting sudden changes in curve slope and peak jitter.

[0057] Harmonic analysis identifies waveform distortion caused by abnormalities or local short circuits in nonlinear components by extracting the harmonic amplitude distribution.

[0058] Based on the comparison results, various characteristic parameters that characterize the electrical properties were extracted from the first voltage-current response curve, including but not limited to:

[0059] Amplitude characteristics: peak voltage, peak current, RMS change, average amplitude offset;

[0060] Phase characteristics: voltage-current phase difference, phase shift amplitude;

[0061] Impedance characteristics: equivalent impedance at different frequencies, slope of impedance change curve;

[0062] Nonlinear characteristics: harmonic amplitude, total harmonic distortion, and energy ratio of specific order harmonics;

[0063] Deformation characteristics: points where the waveform slope changes, number of spikes, and degree of local waveform distortion.

[0064] The above features are combined in a fixed order to form a feature vector, which can fully characterize the electrical state of each test point. The feature vector reflects both absolute value information and relative deviation from the standard curve, thereby improving the sensitivity of anomaly detection.

[0065] The feature vectors of a large number of labeled fault samples are pre-trained offline to form a classification model for multiple fault classes. The model can use structures suitable for fast classification of small samples, such as random forest, support vector machine, or lightweight neural network.

[0066] When the feature vectors of the PCBA under test are input into the model, the model outputs the following structural fault types based on the learned feature patterns:

[0067] Open circuit: The response curve amplitude is close to zero, the phase fluctuation is abnormal, and the impedance shows a sudden increase;

[0068] Short circuit: Voltage response decreases significantly, current rises abnormally, and harmonic characteristics change drastically;

[0069] Component parameter errors: The impedance and phase curves for a specific frequency band deviate, forming a stable difference from the standard curve;

[0070] Leakage current: The low-frequency impedance drops slightly, the current is high under AC excitation but does not reach the short circuit characteristics, and the harmonic components are slightly amplified.

[0071] Finally, the model outputs power failure diagnostic results, including the fault type.

[0072] Specifically, it also includes offline training, including:

[0073] The second electrical signal of key nodes of multiple known normal PCBAs during power-on operation is collected. Based on the second electrical signal, the second feature vector is extracted and a feature database containing only normal data is constructed. An isolated forest model is trained based on the feature database and the trained model is stored locally or in the cloud.

[0074] Understandably, a feature database is constructed using feature vectors collected under normal working conditions, and an isolation forest model is trained based on this feature database. The isolation forest algorithm constructs a large number of independent isolation trees by randomly selecting features and their segmentation thresholds in the feature space. Since the training data all come from normal samples, the model learns during training that normal data points usually require multiple feature segmentations to be completely isolated, meaning their average path length is relatively long.

[0075] During the diagnostic phase, the feature vector corresponding to the PCBA under test is input into the isolated forest model. The model calculates the average path length of this feature vector across all isolated trees and generates an anomaly score accordingly. The anomaly score is typically between 0 and 1, reflecting the similarity between the feature vector and normal samples: when the anomaly score is close to 0.5 or lower, it indicates that its average path length is relatively long, similar to the characteristics of normal data, and can be judged as normal; when the anomaly score is close to 1, it indicates that its average path length is significantly shortened, meaning that the point is easily isolated in the feature space and deviates significantly from normal data, and should be judged as an anomaly.

[0076] In practical applications, when the abnormal score exceeds a preset threshold, the system determines that the PCBA has a dynamic functional abnormality or potential functional failure, such as a shift in signal mean, frequency or other key characteristics due to parameter drift.

[0077] Specifically, based on the historical failure frequency and real-time acquired failure characteristics in the test performance database, the test items that have not yet been executed are dynamically reordered, and the current test sequence is updated, including:

[0078] Before starting the test, the test items are sorted according to the historical failure frequency of each test item recorded in the test performance database to generate an initial test sequence;

[0079] When the current test item is executed and a fault is detected, obtain the fault characteristics corresponding to the current test item;

[0080] Based on the fault characteristics, find one or more test items associated with the fault characteristics in the fault characteristic association graph;

[0081] In the sequence of test items that have not yet been executed, the test item associated with the fault feature is moved to the beginning of the sequence to generate an updated test sequence.

[0082] Specifically, in the test performance database, each test item is assigned a unique name, historical failure frequency, and execution result, and a fault feature association graph is pre-configured so that each type of fault feature corresponds to one or more associated test items.

[0083] It should be noted that in step S7, a two-layer reordering algorithm based on historical frequency and real-time diagnosis is implemented to optimize the test path. Specifically, this includes:

[0084] S71. Establish a test performance database, set a unique name, historical failure frequency, execution result and other fields for each test item, and pre-configure a fault feature association graph so that each type of fault feature corresponds to one or more associated test items.

[0085] S72. Read all test items from the test performance database, sort them according to the historical failure frequency of each test item, and generate an initial test sequence from high to low failure frequency as the first-level test path based on historical frequency.

[0086] S73. Set the current test index to the first test item in the initial test sequence and begin sequential execution of the test process.

[0087] S74. Take the current test item, perform the corresponding test on the PCBA under test, call the power-down diagnostic results and power-on diagnostic results, and output whether there is a fault and the corresponding fault characteristic identifier.

[0088] S75. If no fault is detected in the current test item, the test result is written to the test performance database, the statistics of the test item are updated, and the current test index is incremented by one; if there are still unexecuted test items, return to step S74 to continue executing the next test item.

[0089] S76. If a fault is detected in the current test item, obtain the fault feature identifier of this diagnosis, and query one or more test items associated with it in the fault feature association graph based on the fault feature to obtain a set of associated test items that need to be prioritized.

[0090] S77. In the remaining test sequence after the current test item, find each associated test item in turn, move the found associated test items to the front of the remaining test sequence in turn, while keeping the relative order between the remaining unassociated test items unchanged, thereby reordering the remaining test sequence based on the real-time diagnostic results.

[0091] S78. Concatenate the rearranged remaining test sequence with the executed test item sequence to form an updated global test sequence. Increment the current test index by one, and use the test item at the current index position in the updated sequence as the next test item to be executed. Return to step S74 until all test items have been executed.

[0092] S79. After the entire test process is completed, the actual execution results of each test item, the fault characteristics that occur, and the corresponding reordering adjustment information are written back to the test performance database in a unified manner. The fault frequency statistics of each test item are updated in real time and used to generate new initial test sequences for subsequent batches. This achieves a two-layer test path optimization that combines initial sorting based on historical frequency with dynamic reordering based on real-time diagnosis.

[0093] Example 2

[0094] Based on the same inventive concept, this embodiment provides a PCBA inspection system, referring to... Figure 2 As shown, it includes:

[0095] The power-down diagnostic unit is used to inject AC excitation signals into preset test points on the PCBA when the PCBA is not powered on.

[0096] Based on the AC excitation signal, the first voltage-current response curves corresponding to each test point are acquired in real time.

[0097] Each first voltage-current response curve is compared with its corresponding second voltage-current response curve. Based on the comparison results, feature parameters characterizing electrical characteristics are extracted from each first voltage-current response curve to construct a corresponding feature vector. The feature vector is then input into a pre-trained machine learning classification model to identify faults including open circuit, short circuit, component parameter errors, and leakage, thereby obtaining a power-off diagnosis result. The second voltage-current response curve is a standard voltage-current response curve of the test point under normal conditions stored in a pre-established database.

[0098] The power-on diagnostic unit is used to acquire the first electrical signal of key nodes when the PCBA is powered on and operating; the key nodes include the power rail, clock line, and data bus.

[0099] Statistical features, time-domain features, and frequency-domain features are extracted from the first electrical signal to construct a first feature vector that characterizes the operating state of each key node.

[0100] The first feature vector is input into a pre-trained isolated forest model, anomaly score is calculated based on average path length, and normal or abnormal judgment and corresponding dynamic functional abnormality type are output according to the comparison result of anomaly score and preset threshold, so as to obtain power-on diagnosis result.

[0101] The dynamic reordering unit is used to establish a test performance database that records test item results and fault characteristics based on the power-down diagnostic results and the power-on diagnostic results, and to dynamically reorder the test items that have not yet been executed based on the historical fault frequency and real-time acquired fault characteristics in the test performance database, and update the current test sequence.

[0102] In addition, the PCBA inspection system also includes:

[0103] The graphical user interface module, built on the QT framework, is used to configure the test process, display test results, and provide human-computer interaction.

[0104] The configuration file management module is used to save and load configuration files for test items, test thresholds, and communication parameters in XML and / or JSON format.

[0105] The communication interface module adopts a modular design, supports multiple communication protocols such as UART, RS485, and CAN, and loads the corresponding protocol driver through a plug-in mechanism for data interaction with the PCBA under test and external test hardware.

[0106] The data processing module is used to analyze, store, and generate test reports from the data output by the multimodal electrical fault diagnosis module, and to provide the processing results to the graphical user interface module and the cloud management and analysis platform.

[0107] The cloud management and analysis platform interface module is used to upload local test data to the cloud management and analysis platform, and to receive configuration distribution, diagnostic model updates and risk warning information from the cloud.

[0108] A secure cloud gateway is used to enable encrypted communication, device authentication, and access control between local devices and the cloud platform.

[0109] The remote management interface is used to perform unified configuration management, centralized updates of diagnostic models and test processes, and remote maintenance of multiple PCBA testing systems in the cloud.

[0110] The PCBA testing system adopts a modular architecture, providing a QT-based graphical user interface module, a configuration file management module, a multi-protocol communication interface, a data processing module, a cloud management and analysis platform interface module, and a secure cloud gateway, enabling flexible adaptation to different PCBA testing needs. Through plug-in communication protocols and centralized cloud management, overall maintenance costs are significantly reduced.

[0111] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0112] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0113] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0114] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0115] Finally, it should be noted that the above specific embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to examples, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A PCBA detection method, characterized in that, The method comprises: in a state that the PCBA is not powered on, injecting an alternating excitation signal to a preset test point on the PCBA; based on the alternating excitation signal, collecting a first voltage-current response curve corresponding to each test point in real time; comparing each first voltage-current response curve with a corresponding second voltage-current response curve, extracting a feature parameter for representing an electrical characteristic from each first voltage-current response curve based on a comparison result, constructing a corresponding feature vector, and inputting the feature vector into a pre-trained machine learning classification model to identify a fault including an open circuit, a short circuit, a component parameter error, and a leakage, and obtain a power-off diagnosis result; wherein the second voltage-current response curve is a standard voltage-current response curve of the test point in a normal state stored in a pre-established database; in a state that the PCBA is powered on, collecting a first electrical signal of a key node; the key node includes a power supply rail, a clock line, and a data bus; extracting statistical features, time domain features, and frequency domain features from the first electrical signal to construct a first feature vector for representing a running state of each key node; inputting the first feature vector into a pre-trained isolation forest model, calculating an anomaly score based on an average path length, and outputting a normal or abnormal discrimination and a corresponding dynamic functional abnormality type according to a comparison result of the anomaly score and a preset threshold to obtain a power-on diagnosis result; wherein the isolation forest model constructs independent isolation trees by randomly selecting features and their segmentation thresholds in the feature space; the first feature vector is isolated by feature segmentation of the isolation trees for multiple times to obtain an average path; the isolation forest model calculates the average path length of the first feature vector in all isolation trees and generates an anomaly score based thereon, the anomaly score being between 0 and 1 and reflecting the similarity of the feature vector to normal samples; based on the power-off diagnosis result and the power-on diagnosis result, establishing a test performance database recording test item results and fault characteristics, and dynamically reordering unexecuted test items based on historical fault frequencies in the test performance database and real-time acquired fault characteristics to update a current test sequence; wherein dynamically reordering unexecuted test items based on historical fault frequencies in the test performance database and real-time acquired fault characteristics to update a current test sequence comprises: sorting test items according to historical fault frequencies of each test item recorded in the test performance database before starting the test to generate an initial test sequence; when a fault is detected during execution of a current test item, acquiring a fault characteristic corresponding to the current test item; finding one or more test items associated with the fault characteristic in a fault characteristic association graph according to the fault characteristic; moving the test items associated with the fault characteristic to the front of the sequence in the sequence of test items that have not yet been executed to generate an updated test sequence.

2. The PCBA detection method of claim 1, wherein, the statistical features include mean, variance, standard deviation, peak-to-peak value, skewness, and kurtosis of the signal; the time domain features include rise time, fall time, pulse width, and duty cycle; and the frequency domain features include frequency, amplitude, and phase. The frequency domain features include a main frequency, harmonic component energy and total spectral energy obtained based on a fast Fourier transform.

3. The PCBA detection method of claim 1, wherein, The offline training comprises: The second electric signals of the key nodes of the known normal multiple PCBAs during power-on operation are collected, the second feature vectors are extracted based on the second electric signals, a feature database containing only normal data is constructed, an isolation forest model is trained based on the feature database, and the trained model is stored locally or in the cloud.

4. The PCBA detection method of claim 1, wherein, In the test performance database, a unique name, a historical failure frequency, an execution result are set for each test item, and a failure feature correlation graph is pre-configured, so that each type of failure feature corresponds to one or more associated test items.

5. A PCBA detection system characterized by, The detection system for implementing the PCBA detection method of any one of claims 1-4 comprises: A power-off diagnosis unit configured to inject an alternating excitation signal to a preset test point of the PCBA in a power-off state; Real-time collection of first voltage-current response curves corresponding to each test point based on the alternating excitation signal; Comparison of each first voltage-current response curve with a corresponding second voltage-current response curve, extraction of feature parameters for representing electrical characteristics from each first voltage-current response curve based on the comparison result, construction of a corresponding feature vector, input of the feature vector into a pre-trained machine learning classification model, identification of failures including open circuit, short circuit, component parameter error and electric leakage, and obtaining of a power-off diagnosis result; wherein the second voltage-current response curve is a standard voltage-current response curve of the test point in a normal state stored in a pre-established database; A power-on diagnosis unit configured to collect first electric signals of key nodes in a power-on working state of the PCBA; the key nodes include power supply rails, clock lines and data buses; Extraction of statistical features, time domain features and frequency domain features from the first electric signals, and construction of a first feature vector for representing the running states of each key node; Input of the first feature vector into a pre-trained isolation forest model, calculation of an anomaly score based on an average path length, output of normal or abnormal discrimination and a corresponding dynamic functional abnormality type according to a comparison result of the anomaly score and a preset threshold, and obtaining of a power-on diagnosis result; A dynamic reordering unit configured to establish a test performance database recording test item results and failure features based on the power-off diagnosis result and the power-on diagnosis result, to dynamically reorder unexecuted test items based on a historical failure frequency in the test performance database and real-time acquired failure features, and to update a current test sequence.

Images