Partition detection method for board card quality

Abstract

The invention discloses a partition detection method for board card quality, and relates to the technical field of detection. The method comprises the following steps: dividing a board card into functional blocks according to a functional structure, and collecting thermal field distribution characteristic data of each block; constructing a standard thermal template, calculating the thermal imaging matching degree of the functional blocks by using an Euclidean distance algorithm, and marking the thermal imaging matching degree exceeding a preset threshold value as an abnormal thermal response region; collecting resistance and resistance gradient change data of an abnormal thermal response area, judging the area as a defect cluster area if conditions are met, and calculating a resistance abnormal amplitude; integrating defect cluster region data to form a composite characteristic spectrum, inputting a defect sample matching model, and outputting a defect classification result and a confidence score; calculating a regional quality score by combining fuzzy logic reasoning and a neural network algorithm and integrating a thermal imaging matching degree, a resistance abnormal amplitude and a confidence score; and fusing the defect classification result and the regional quality score to generate a quality grade label. And based on the quality grade labels, board card quality partition detection is completed.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of detection engineering, and particularly to a partition detection method for board card quality. BACKGROUND

[0002] In today's highly integrated electronic devices, the quality of the board card as a key component of the electronic system is directly related to the performance and reliability of the electronic device. Whether it is the motherboard, graphics card, network card of the computer, or the circuit board of the communication device and industrial control system, it carries the key signal processing, data transmission and control functions. Any minor defect may cause partial or even overall functional failure, resulting in device failure, loss and inconvenience to users.

[0003] However, the traditional board card quality detection relies on separate thermal imaging detection, which can only find thermal abnormal areas, but cannot determine the specific reasons for the thermal abnormalities, and it is difficult to accurately locate and classify hidden defects. SUMMARY

[0004] In view of the deficiencies of the prior art, the present application provides a partition detection method for board card quality, which solves the problem that the traditional method is difficult to accurately locate and classify hidden defects.

[0005] To achieve the above purpose, the present application realizes the following technical scheme: a partition detection method for board card quality, comprising the following steps: Step S1: dividing the to-be-tested board card according to the functional structure to obtain board card functional blocks; collecting thermal imaging of the board card functional blocks to obtain thermal field distribution characteristic data; Step S2: constructing standard thermal template data, quantifying the difference between the thermal field distribution characteristic data of each board card functional block and the standard thermal template data by using the z-score maximum value method, and calculating the thermal imaging matching degree of each functional block; when the thermal imaging matching degree of the board card functional block is greater than a preset threshold, the board card functional block is marked as an abnormal thermal response area; Step S3: collecting resistance data and resistance gradient change data of the abnormal thermal response area; when the resistance value data or the resistance gradient change data meet the preset defect cluster area locking condition, the abnormal thermal response area is further determined as a defect cluster area, and the resistance abnormal amplitude is calculated based on the resistance value and the resistance gradient change data of the abnormal thermal response area; Step S4: combining the thermal field distribution characteristic data, the resistance data and the resistance gradient change data of the defect cluster area to obtain a composite feature map; constructing a defect sample matching model, inputting the composite feature map into the defect sample matching model, and outputting to obtain a defect classification result and a confidence score; Step S5: According to a fuzzy logic reasoning and neural network combined algorithm, the thermal imaging matching degree, the resistance anomaly amplitude and the confidence score are combined to calculate the regional quality score; Step S6: combining the defect classification result with the regional quality score to generate a detection region quality grade label, and implementing zoning detection of board quality based on the regional quality grade label.

[0006] Preferably, the step S1 includes: Based on the core principles of functional independence and signal flow of the board, combined with the modular design in the circuit diagram, the circuit units that perform the same function are divided into a block, including: power module, main control module, interface module and power load module; Each block corresponds to a thermal imaging unit that can be analyzed independently. The infrared thermal imager performs multi-state scanning on the powered board to collect temperature data of each functional block. Based on the temperature data of each functional block, the thermal field distribution cloud map, temperature rise curve and hot spot morphological characteristics are generated.

[0007] Preferably, the construction of standard thermal template data includes: Standard thermal template data construction: Collect thermal imaging data of qualified boards under the same working conditions, remove outliers and calculate the statistical feature data of each block: Temperature mean = the average temperature of the block of all qualified boards; Temperature standard deviation = reflects the temperature fluctuation range between qualified boards; Hot spot morphological characteristics: including the area, shape factor, and centroid position of normal hot spots; Template update mechanism: After every 500 boards are inspected, new qualified data is automatically incorporated to recalculate μ and σ to adapt to process fluctuations.

[0008] Preferably, the z-score maximum value method is used to quantify the difference between the thermal field distribution characteristic data of each functional block and the standard thermal template data, and the thermal imaging matching degree of each functional block is calculated to obtain the following: The difference between the measured data and the standard thermal template data is quantified by the z-score maximum method to obtain the thermal imaging matching degree. The formula is as follows: Calculate the z-score of each point:

[0009] in, It is The temperature measurement value of each point; It is The mean temperature of each point; It is The standard deviation of the temperature at each point; Among all the calculated z-scores, take the maximum value : ; Define the degree of match:

[0010] in, is the degree of thermal imaging matching, k is the attenuation coefficient, which determines the attenuation speed, is the point in the data set that deviates most from the mean. Range ∈(0,1], The higher the value, the closer the measured thermal field is to the standard state. ≥0.9: good match, no significant thermal anomaly; 0.7< <0.9: Slight deviation, needs to be analyzed in combination with other dimensions; ≤0.7: Significant abnormality, marked as "abnormal thermal response area".

[0011] Preferably, collecting the resistance data and resistance gradient change data of the abnormal thermal response area includes: Contact probe: Suitable for high-precision detection. The probe tip directly contacts the surface of the object being measured to accurately collect surface resistance or electrical signal changes. Precision mechanical contact ensures detection stability and reliability. Meshing: Divide the abnormal thermal response area into two-dimensional grids with a spacing of 0.5 mm. For example, a 10 mm × 10 mm area is divided into 20 × 20 = 400 grid points. Each grid point is marked as (i, j), corresponding to the board coordinates ( , ); Point-by-point scanning process: The probe positioning servo motor drives the probe array to move to just above point (i, j), and the contact probe applies a pressure of 10 g; For resistance measurement, apply a 100mV excitation voltage, measure the current flowing through the probe, and calculate the surface resistance using Ohm's law:

[0012] in, is the resistance value at the (i, j) grid point, V is the input voltage, and I is the resistance value; Gradient calculation of the resistance gradient between the current point and the four adjacent points:

[0013] in, Indicates the current point resistance gradient; Represents the resistance value of the current point; 、 Respectively represent the resistance values ​​of the adjacent points above and below the current point; 、 They represent the resistance values ​​of the adjacent points to the right and left of the current point respectively. The gradient value reflects the degree of mutation in the resistance distribution. The larger the gradient, the higher the possibility of defects.

[0014] Preferably, the calculation of the resistance abnormality amplitude based on the resistance value and resistance gradient change data of the abnormal thermal response area includes: The regional resistance anomaly amplitude is calculated by using the RMS weighted deviation method. The formula is as follows:

[0015] in, is the resistance anomaly amplitude, It is the first The resistance value of each abnormal grid point; is the average normal resistance of the area; is the weight coefficient, which is set inversely proportional to the distance between the grid point and the cluster center; is the total number of abnormal grid points; the numerator is the weighted sum of the squares of the differences between the resistance values ​​of each abnormal grid point in the defect cluster and the average normal resistance value in that area. The denominator is the sum of the weight coefficients of all abnormal grid points. This calculation method uses the RMS weighted deviation method to calculate the resistance anomaly amplitude of the cluster area, more accurately assessing the resistance anomaly by considering the weight of each abnormal grid point.

[0016] Preferably, the composite characteristic map obtained by combining the thermal field distribution characteristic data, resistance data and resistance gradient change data of the defect cluster region comprises: The following key features are extracted from the defect cluster area to construct a composite feature map: Among them, thermal imaging features include: Hot spot centroid coordinates: ( , ), reflecting the location of the hot spot on the board; the maximum value of the thermal gradient: , indicating the most intense degree of thermal field change in the region; Average temperature: , reflecting the overall fever level of the area; Resistance sweep features include: Average resistance value: , which measures the average level of electrical conductivity in the area; Resistance gradient mean: , describes the average degree of change in resistance value in the spatial distribution; Impedance reversal times: , records the number of times "high-low-high" or "low-high-low" resistance fluctuations occur in the region; Defect cluster area: , represents the size of the abnormal resistance area; The above features are integrated into a composite feature map : ; The composite feature map can be represented as a vector in practical applications. This representation is consistent with the relevant definitions and data processing logic. From a definitional perspective, the map is essentially an integrated representation of multiple feature information, and the vector is a high-efficiency data structure that can store and operate on these features in an ordered numerical form.

[0017] Preferably, the output obtains defect classification results and confidence scores, including: The cosine similarity algorithm is used to calculate the similarity between the feature vector of the sample to be detected and the feature vector of the historical sample : :

[0018] Wherein, is the similarity between the feature vector of the detection sample and the feature vector of the historical sample, is the feature vector of the detection sample, is the feature vector of the historical sample, The value range of , the value is closer to , indicating that the two vectors are more similar. The numerator is the calculation formula of the dot product of the two vectors, representing the sum of the products of the corresponding elements in the feature vectors of the detection sample and the historical sample, is the jth feature vector in the feature vector of the detection sample, is the jth feature vector in the feature vector of the historical sample; Defect type output and confidence score: Select the top historical samples with the highest similarity, sort them from high to low according to the similarity, and the predicted defect type is the defect type corresponding to the sample with the highest similarity , the confidence score is converted from the highest similarity : ; Wherein, is the confidence score, ​is the highest similarity.

[0019] Preferably, the area quality score is calculated by combining the thermal imaging matching degree, the resistance anomaly amplitude and the confidence score according to the fuzzy logic reasoning and neural network combination algorithm, including: Nonlinear construction of regional quality scoring function: By combining fuzzy logic reasoning with neural networks, the thermal imaging matching degree, resistance anomaly amplitude, and confidence score are converted into regional quality scores. Among them, the fuzzy logic preprocessing is: Define fuzzy sets: Map raw data to "low / medium / high" fuzzy levels: The thermal imaging matching degree is: Low: ,middle: ,high: ; The abnormal resistance amplitude is: Low ),middle ,high: ; The confidence scores are: Low: ,middle: ,high: ; Establish a fuzzy rule base: formulate rules through historical data training or expert experience, as follows:

[0020] Among them, the neural network includes: Input layer: takes the fuzzified features × 3 dimensions and the original confidence scores as input; Hidden layer: uses 2 fully connected layers and ReLU as the activation function; The first fully connected layer includes: Construct the weight matrix: ; Dimension meaning: number of input features × number of neurons Parameter function: each weight Indicates the The input feature pair The influence strength of each neuron; The calculation process is: =ReLU( · + ); in, is the first layer output vector, the feature representation after ReLU activation, Y is a 10-dimensional input vector, containing fuzzed features and original confidence, is a 16-dimensional bias vector, used to adjust the neuron activation threshold, ReLU is the activation function: introducing nonlinearity, the formula is ReLU( )=max(0, ), filtering negative signals; The second layer is a fully connected layer, which includes: The weight matrix: ; Dimension meaning: 16 first layer output number x 8 second layer neuron number Parameter role: compress the high-dimensional features extracted by the first layer into more abstract representations The calculation process is: ; Among them, The second layer output vector, the final abstract feature representation, is the 16-dimensional vector of the first layer output, is an 8-dimensional bias vector, is the activation function, used to introduce nonlinearity; Output layer: single node output area quality score , mapped to the score interval by the Sigmoid function; Training method: using historical quality data, using Adam optimizer to minimize the mean square error loss function.

[0021] Preferably, the combination of the defect classification result and the area quality score to generate a detection area quality level label comprises: The quality level label is determined by a decision tree, specifically: A CART decision tree model is constructed, taking the area quality score and the defect type as core features, and outputting the quality level label ; The decision tree is constructed as follows: Splitting condition: First, divide according to the area quality score: Stable area; Further determine the defect type; Potential defect area or fault area; Second, refine according to the defect type: If the defect type is "fatal defect", it is directly determined as fault area; if it is "repairable defect", the level is adjusted in combination with the score; Pruning optimization: avoid overfitting by cross-validation, set the minimum sample size to 10, and the maximum depth to 5; Decision path: Input: , "virtual soldering of soldering points" -> decision tree determines slight abnormal area; input: , "power layer short circuit" -> decision tree determines fault area.

[0022] Beneficial effects: The application provides a partition detection method for board quality, which relates to machine learning and deep learning technology, and has the following beneficial effects: (1) The partition detection method for board quality combines thermal imaging acquisition and resistance data monitoring, uses Euclidean distance algorithm to quantify thermal field difference, accurately locates abnormal thermal response area, and further determines defect cluster area, effectively improves detection accuracy, can find small thermal anomalies and resistance changes, and reduces the risk of missing potential defects; (2) The partition detection method for board quality innovatively fuses thermal field distribution feature data, resistance data and resistance gradient change data, constructs a composite feature map, fully excavates the information value of multi-source data, provides more abundant and comprehensive basis for board quality evaluation, and makes the detection result more reliable and credible; (3) The partition detection method for board quality combines thermal imaging matching degree, resistance abnormal amplitude and confidence score, constructs regional quality score, and combines with defect classification result to generate regional quality level label, realizes comprehensive and comprehensive evaluation of board quality, and more accurately reflects the actual quality status of the board. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1 A flowchart of the partition detection method for board quality is provided.

[0024] Figure 2 A hierarchical diagram of the partition detection method for board quality is provided. DETAILED DESCRIPTION

[0025] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, not all. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the application.

[0026] Please refer to Figure 1The application provides a technical scheme: a partition detection method for board quality. Specifically, the following board quality partition detection method is provided. Please refer to Figure 1 The method comprises the following steps: Step S1: divide the to-be-detected board according to the functional structure to obtain a board functional block; and collect the thermal imaging of the board functional block to obtain thermal field distribution characteristic data.

[0027] Based on the circuit design principle and physical layout characteristics of the board, the complex board system is disassembled into independent blocks with clear functional properties, and a mapping relationship between “function-space” is established, thereby providing a structured analysis unit for subsequent thermal imaging detection. This process needs to deeply integrate circuit knowledge, component characteristics and detection requirements, and the specific operation is as follows: Firstly, the division basis is determined: taking the functional independence and signal flow direction of the board as the core principle, combining the modular design in the circuit diagram (such as power supply module, main control module, interface module, etc.), the circuit units completing the same type of function are divided into a block: The power supply and voltage stabilization module: including power input interface, filter capacitor, voltage stabilizer chip, inductor and other components, responsible for converting external power supply into stable voltage required by each module of the board, with an independent power supply circuit.

[0028] The signal processing module: covering the central processing unit (CPU), digital signal processor (DSP), storage chip and peripheral resistance / capacitance network, responsible for data operation, logic control and other core functions, and is the “brain” of the board.

[0029] The communication interface module: including USB interface, Ethernet interface, radio frequency antenna interface and the like, responsible for data transmission or signal interaction with external devices, and usually has an independent signal isolation and impedance matching circuit.

[0030] The power load module: including power transistor, relay, motor drive chip and the like, used for driving large current load (such as motor, electromagnetic valve), with high power consumption, and the heat generation characteristics are significantly different from other modules Secondly, precise division is realized based on the design document: Analyze the circuit diagram: by reading the schematic diagram, identify the circuit boundaries of each functional module (such as isolation components between the power supply module and the signal module, distribution of ground symbols), and mark the key components of each module (such as the VCC pin of the power supply chip, the clock pin of the signal chip).

[0031] Compare the BOM list: according to the model and parameters of the components in the bill of materials, judge the functional type of the components (such as electrolytic capacitor is commonly used for power supply filtering, high-speed operational amplifier belongs to the signal processing module), and record the coordinate position of the components on the PCB.

[0032] Finally, each functional block must meet the following criteria: Functional singularity: all elements within the block serve the same core function, with no redundant or unrelated elements; Physical divisibility: the block has clear spatial boundaries on the PCB (distinguishable through solder mask, isolation slots, or element arrangement spacing); Detection convenience: the block size is moderate (recommended not to exceed 100mm 2 ), facilitating thermal imaging focusing and scanning, and each block contains at least 1-2 characteristic elements (such as the heat sink of the power supply module, the main control chip of the signal module) as reference points for thermal field analysis.

[0033] Through the above steps, the board card is transformed into a "detection map" composed of multiple functional blocks, each block corresponding to an independently analyzable thermal imaging unit, laying the foundation for subsequent defect positioning based on functional semantics. This division not only improves the detection specificity (such as directly locating the power supply module overheating problem), but also provides a logical framework for cross-block linkage analysis (such as the impact of power supply abnormalities on signal modules).

[0034] Obtain the thermal field distribution of each block, the core operation: use an infrared thermal imager to perform multi-state scanning on the powered board card, collect temperature data of each functional block, and construct a dynamic thermal field distribution model to provide original physical signals for abnormality determination. The specific process is as follows: Device parameter configuration and environmental control Thermal imager settings: Resolution: ≥640x480 pixels, ensuring that the temperature of individual components (such as 0402 packaged resistors) can be distinguished; Frame rate: 10 frames / second, capturing rapid changes in thermal response (such as temperature fluctuations under pulsed load); Wavelength range: 7.5-14μm (suitable for long-wave infrared detection of normal temperature objects).

[0035] Environmental requirements: Detection chamber temperature: 25℃±1℃, humidity: 45%-55%RH, to avoid environmental temperature and humidity gradients interfering with board card heat dissipation; Electromagnetic shielding: turn off nearby strong electromagnetic equipment (such as frequency converters, motors) to prevent electromagnetic interference affecting the accuracy of the thermal imager sensor.

[0036] Multi-condition thermal imaging data collection: Idle state (static power consumption detection): power on the board card but do not connect the load (such as disconnect the signal input, run the power supply module under no load), and scan after 5 minutes of stabilization; Objective: detect abnormal heating in standby state (such as excessive static current of power supply chip, leakage of energy storage capacitor).

[0037] Full load state (extreme working condition detection): the board runs at maximum load (e.g., the signal processing module performs full load operation, and the power module drives the rated load), and runs for 30 minutes to reach thermal equilibrium; Objective: to expose design defects of heat dissipation (e.g., poor contact between chip and heat sink, insufficient current carrying capacity of PCB copper foil).

[0038] Thermal field distribution characteristic data record: Each block generates a thermal field distribution cloud chart (temperature chromatogram, with the highest / lowest temperature points marked), a temperature rise curve (temperature change trend over time), and thermal spot morphological characteristics (area, centroid position coordinates, etc.).

[0039] Among them, thermal spot is a phenomenon of abnormal temperature rise in a local area of the board due to high current density or poor contact, usually appearing as a local high temperature area in the thermal imaging image, Thermal spot morphological characteristics acquisition method: Area: convert the actual area by counting the number of pixel points covered by the thermal spot in the thermal imaging image and combining the pixel-actual size conversion ratio of the infrared device ; Centroid position coordinates: calculate the average of each pixel coordinate in the thermal spot area by formula:

[0040] Among them, and are the coordinates of the centroid position on the x-axis and y-axis, is the coordinate of each pixel in the thermal spot area, is the total number of pixels in the thermal spot area.

[0041] This step divides the board into functional blocks according to functional structures (e.g., power module, main control module), and collects the thermal field distribution characteristics of each block by infrared thermal imager. This way, complex boards can be decomposed into structured units, reducing analysis complexity. The thermal field data obtained provides the original basis for subsequent comparison with standard templates and positioning of abnormal areas.

[0042] Step S2: Construct standard thermal template data, use z-score maximum value method to quantify the difference between each functional block thermal field distribution characteristic data and standard thermal template data, and calculate the thermal imaging matching degree of each functional block. When the thermal imaging matching degree of the board functional block is greater than the preset threshold, the board functional block is marked as an abnormal thermal response area.

[0043] Standard thermal template data construction: collect thermal imaging data of qualified boards under the same working conditions, and calculate the statistical characteristic data of each block after removing outliers: Temperature mean (μ) = temperature average of all qualified boards in this block; Temperature standard deviation (σ) = reflects the temperature fluctuation range between qualified boards; Hot spot morphological characteristics: normal hot spot area, centroid position.

[0044] Template update mechanism: after detecting 500 boards, automatically include new qualified data to recalculate μ and σ, adapt to process fluctuations (such as component batch differences, soldering process adjustment).

[0045] Quantify the difference between the measured data and the standard thermal template data by z-score maximum method, get the thermal imaging matching degree, the formula is as follows: Calculate the z-score of each point:

[0046] Among them, is the temperature measurement value of the th point; is the temperature mean value of the th point; is the temperature standard deviation of the th point.

[0047] Among all the calculated z-scores, take the maximum value : ; Define the matching degree:

[0048] Among them, is the thermal imaging matching degree, k is the attenuation coefficient, which determines the speed of attenuation, is the point most deviating from the mean value in the data set, value range ∈(0,1], when it automatically approaches 0 (consistent with the 3σ principle); the higher the value, the closer the measured thermal field to the standard state: ≥0.9: good match, no significant thermal anomaly; 0.7 <0.9: slightly deviate, need to be combined with other dimension analysis; ≤0.7: significant anomaly, marked as "abnormal thermal response area".

[0049] On the basis of matching degree quantitative analysis, superimpose the following qualitative judgment rules: Absolute temperature limit value: if the temperature of any point in the block exceeds the safety threshold (such as the maximum junction temperature limit of the component-10℃), it is directly determined as abnormal, without calculating the matching degree.

[0050] Hot spot morphological anomaly: hot spot area exceeds 2 times the standard value (such as standard hot spot area ≤50mm 2 , measured >100mm2 The matching degree calculation result is combined with the qualitative rule: the matching degree is less than or equal to 0.7 or a block or hot spot shape anomaly triggering an absolute temperature limit value is marked as an "abnormal heat response area" and the coordinate range of the "abnormal heat response area" is labeled.

[0051] This step is based on the thermal imaging data of qualified board cards to construct a standard thermal template (including temperature mean value, standard deviation, hot spot shape, etc.), and the difference between the measured thermal field and the standard template is calculated to obtain the thermal imaging matching degree. The abnormal heat response area is marked when the matching degree exceeds the threshold value (such as ≤0.7). Through data-driven positioning of the thermal field abnormal area, invalid detection is reduced; the quantification of the matching degree (0-1 interval) provides a comparable index for the abnormality degree.

[0052] Step S3: Collecting resistance data and resistance gradient change data of the abnormal heat response area, when the resistance value data or resistance gradient change data meet the preset defect cluster area locking condition, further determining the abnormal heat response area as a defect cluster area, and calculating the resistance abnormal amplitude based on the resistance value and resistance gradient change data of the abnormal heat response area.

[0053] The labeled "abnormal heat response area" is scanned by resistance, and the thermal anomaly is verified whether it is caused by a real physical defect (such as a virtual weld or a crack) through electrical characteristics, avoiding the waste of resources caused by full-area blind scanning, and realizing "thermal-electric" double verification.

[0054] Selection of detection equipment: Contact probe: suitable for high-precision detection (such as 0.1mm pitch solder joints), through direct contact between the probe tip and the surface of the measured object, accurate collection of surface resistance or electrical signal changes, and precise mechanical contact to ensure detection stability and reliability.

[0055] Grid division: Divide the two-dimensional grid in the abnormal heat response area by 0.5mm interval, for example, divide the 10mm×10mm area into 20×20=400 grid points.

[0056] Each grid point is marked as (i,j), corresponding to the board card coordinates ( , ).

[0057] Point-by-point scanning process: The probe positioning servo motor drives the probe array to move to the (i,j) point directly above, and the contact probe applies 10g pressure; Resistance measurement applies 100mV excitation voltage (to avoid damaging the element), measures the current value flowing through the probe, and calculates the surface resistance value through Ohm's law: ​

[0058] wherein, R(i,j) is the resistance value at the (i,j) grid point, V is the input voltage, and I is the current.

[0059] The gradient of the current point and its adjacent four points (up, down, left, right) is calculated:

[0060] wherein, represents the resistance gradient of the current point ; represents the resistance value of the current point; , and R(i-1,j) and R(i+1,j) represent the resistance values of the adjacent points to the left and right of the current point, respectively; , and R(i,j-1) and R(i,j+1) represent the resistance values of the adjacent points above and below the current point, respectively. The gradient value reflects the degree of sudden change in the resistance distribution, and the greater the gradient, the higher the possibility of defects.

[0061] Locking conditions for defect cluster areas: When any of the following conditions is met in a certain area, it is determined to be a "defect cluster area": Impedance reversal occurs in 5 or more adjacent grid points: the resistance values of adjacent grid points show "high-low-high" or "low-high-low" fluctuations, and the fluctuation amplitude exceeds 2 times the normal mean value (e.g., if the normal resistance value is 50 mΩ, the fluctuation range is > 100 mΩ or < 25 mΩ).

[0062] Resistance gradient of 5 or more adjacent grid points exceeds the limit: the resistance gradient > 50 mΩ / mm (normal conductive area gradient < 10 mΩ / mm), and the adjacent grid points are continuously distributed in space (e.g., forming a linear or planar area).

[0063] For the locked "defect cluster area", first calculate the area of the defect cluster area, and then further calculate the overall amplitude level of the resistance anomaly in the area to quantify the severity of the defect. The specific steps are as follows: Calculate the area S of the defect cluster area: Single grid area: the physical area of each grid is 0.5 mm x 0.5 mm = 0.25 m ; Abnormal point quantity statistics: count the total number of abnormal grid points (denoted as q) that meet the locking conditions of the defect cluster area.

[0064] Total area calculation: S = q x 0.25 m ; Wherein, S is the area of the defect cluster region, and q is the total number of abnormal grid points meeting the locking condition of the defect cluster region.

[0065] The area resistance anomaly amplitude calculation adopts the RMS weighted deviation method to calculate the resistance anomaly amplitude of the cluster area, and the formula is as follows:

[0066] Wherein, is the resistance anomaly amplitude, is the resistance value of the i-th abnormal grid point in the defect cluster region; is the normal resistance mean value of the region (taken from the standard template); is the weight coefficient, which is inversely proportional to the distance between the grid point and the cluster center (the closer the distance, the higher the weight, and the maximum value is 1); is the total number of abnormal grid points (q).The physical meaning of the numerator is the weighted sum of squares of the resistance values of each abnormal grid point in the defect cluster region and the difference between the normal resistance mean value of the region. The physical meaning of the denominator is the total weight coefficient of all abnormal grid points. This calculation method adopts the RMS weighted deviation method to calculate the resistance anomaly amplitude of the cluster area, and more accurately evaluates the resistance anomaly condition by considering the weight of each abnormal grid point. This step divides the grid for the abnormal thermal response area, collects the resistance value and resistance gradient of each grid point through probe scanning, and when the resistance or gradient meets the defect locking condition (such as the resistance gradient of more than 5 adjacent points exceeding the limit), it is determined as a defect cluster region, and the resistance anomaly amplitude is calculated by the RMS weighted deviation method. Through the “thermal-electric” double verification to lock the real defect area, the false judgment of thermal anomaly is excluded; the resistance anomaly amplitude quantifies the electrical severity of the defect, which provides a basis for subsequent quality scoring.

[0067] Step S4: combining the thermal field distribution characteristic data, resistance data and resistance gradient change data of the defect cluster region to obtain a composite characteristic spectrum; constructing a defect sample matching model, inputting the composite characteristic spectrum into the defect sample matching model, and outputting to obtain a defect classification result and a confidence score.

[0068] Composite characteristic spectrum construction: The following key features are extracted from the defect cluster area to construct a composite characteristic spectrum:

[0069] Wherein, the thermal imaging features include: The coordinates of the thermal spot centroid: (x, y), reflecting the position of the thermal spot on the board card; the maximum thermal gradient: ​​​, which represents the most severe degree of thermal field change in the region, the maximum value of the absolute value of the temperature change rate (based on the slope change in the temperature rise curve); Average temperature: , which reflects the overall heating level of the region.

[0070] Resistance scanning features include: Average resistance value: , which measures the average level of electrical conductivity in the region; Resistance gradient mean: , which describes the average degree of change in resistance value in spatial distribution; Impedance inversion times: , which records the number of "high-low-high" or "low-high-low" resistance fluctuations in the region.

[0071] Defect cluster area: , which represents the size of the abnormal resistance area.

[0072] The above features are integrated into a real-time feature vector (composite feature map) : ; The composite feature map can be represented as a vector in practical applications. This representation is consistent with the relevant definitions and data processing logic. From the definition point of view, the map is essentially an integrated expression of multiple feature information, while the vector is a high-efficiency data structure that can store and operate these features in an ordered numerical form.

[0073] Defect sample matching model construction: First, a large number of historical defect sample data need to be collected. For each sample, record its defect type (such as tin bead short circuit, chip virtual welding, etc.), and through the detection process of steps S1 and S3, obtain the thermal imaging features (thermal spot centroid coordinates (x, y) , ), maximum thermal gradient , average temperature , etc.) and resistance scanning features (average resistance value , resistance gradient mean , impedance inversion times , defect cluster area , etc.) corresponding to the sample. Organize these features into historical feature vectors, and perform Z-Score standardization on the historical feature vectors to obtain standardized historical feature vectors , where , is the total number of defect samples. At the same time, record the defect type label of each sample, and perform Z-Score standardization on the detection sample feature vector Z-Score standardization is also performed to keep the dimension consistent.

[0074] Feature weight determination: Different features have different importance in defect type recognition. In order to reflect this difference in importance, it is necessary to determine the feature weight vector Based on the expert experience method, experts in the field are invited to score the importance of each feature based on past repair and detection experience, and then normalized to obtain the weight; Suppose we invite experts to score the importance of features (such as hot spot centroid, resistance gradient, etc.), and each expert gives a score of ( is the expert number, is the feature number), and the score range is usually 1-10 points (10 points are the most important).

[0075] Single feature total score calculation, for each feature , calculate the sum of all expert scores: ; Where, is the sum of expert scores of the jth feature, i.e. the sum of all expert scores for this feature, the total number of experts participating in scoring; k is the expert number, the value range ( ), is the score of the kth expert for the jth feature.

[0076] Feature weight normalization: Convert the single feature total score to weight , to ensure that the sum of all weights is 1: ; Where, the weight value of the jth feature reflects the relative importance of the feature in defect recognition; is the sum of expert scores of the jth feature; n is the total dimension of the feature (representing 7 features such as hot spot position, resistance response), and j is the feature number.

[0077] Weighted processing is performed on the feature vector to obtain the weighted feature vector and : ; ; Where, is the detection sample feature vector, which is the weighted feature vector of the real-time feature vector X, is the historical sample feature vector, is the historical feature vector The weighted eigenvector.

[0078] Use the cosine similarity algorithm to calculate the feature vector of the sample to be tested and historical sample feature vector Similarity :

[0079] in, is the similarity between the detection sample feature and the historical sample feature vector, i is the sample index, is the detection sample feature vector, is the historical sample feature vector, The value range is , the closer the value is to , which means the two vectors are more similar. The numerator is the calculation formula of the dot product of two vectors, which means the sum of the products of the corresponding elements in the feature vectors of the sample to be tested and the historical sample. is the jth feature vector in the detection sample feature vector, is the jth eigenvector in the historical sample eigenvector.

[0080] Defect type output and confidence score: Select the one with the highest similarity Historical samples ( Usually 3 is selected) and the similarity is sorted from high to low. The predicted defect type is the defect type corresponding to the sample with the highest similarity. , confidence score By the highest similarity Convert to:

[0081] in, is the confidence score, is the highest similarity. For example, if , then the confidence score If the highest similarity is lower than the set threshold (such as ), it is determined to be an unknown defect type and marked as "unrecognizable".

[0082] The step is to fuse the thermal field features (thermal spot centroid, thermal gradient maximum, etc.) and resistance features (average resistance, resistance gradient mean, etc.) of the defect cluster area to construct a composite feature map, input the defect sample matching model (based on cosine similarity and historical sample comparison), output the defect type (such as false welding, short circuit) and confidence score, and comprehensively improve the defect recognition accuracy by combining thermal and electrical characteristics; the classification result clearly defines the defect type, and the confidence score quantitatively judges the reliability, providing a basis for subsequent decision-making.

[0083] Step S5: According to the fuzzy logic reasoning and neural network combination algorithm, the thermal imaging matching degree, resistance abnormal amplitude and confidence score are combined, and the regional quality score is calculated.

[0084] Nonlinear construction of regional quality score function: The fuzzy logic reasoning and neural network combination method is used to convert the thermal imaging matching degree ( ), resistance abnormal amplitude ( ), confidence score ( ) into regional quality score ( ). Among them, the fuzzy logic preprocessing is: Define fuzzy set: map the original data to "mild / medium / severe" fuzzy levels. For example: The thermal imaging matching degree is: Low ( ), medium ( ), high ( ). The resistance abnormal amplitude is: Low ( ), medium ( ), high ( ).

[0085] The confidence score is: Low ( %), medium ( ), high ( ).

[0086] Establish a fuzzy rule base: develop rules through historical data training or expert experience, for example:

[0087] Among them, the neural network includes: Input layer: fuzzy features (3 fuzzy level vectors) x 3 dimensions and original confidence score as input, where the original confidence score is based on historical data training or expert experience, and the reliability of the data is considered; Hidden layer: 2 fully connected layers (16, 8 neurons respectively), with ReLU as the activation function.

[0088] First fully connected layer (16 neurons): Constructing the weight matrix: ; Dimensionality: 10 (number of input features) x 16 (number of neurons) Parameter role: Each weight represents the influence strength of the th input feature on the th neuron; Its calculation process is: = ReLU( · + ); Where, is the first layer output vector, the feature representation after ReLU activation, Y is a 10-dimensional input vector containing fuzzy features and original confidence, is a 16-dimensional bias vector used to adjust the neuron activation threshold, and ReLU is the activation function: introducing nonlinearity, the formula is ReLU( )= max(0, ), filtering negative signals.

[0089] Second fully connected layer (8 neurons): Weight matrix: ; Dimensionality: 16 (first layer output number) x 8 (second layer neuron number) Parameter role: compress the high-dimensional features extracted by the first layer into more abstract representations Its calculation process is: ; Where, is the second layer output vector, the final abstract feature representation (8-dimensional), is the 16-dimensional output of the first layer, is an 8-dimensional bias vector, is the activation function, used to introduce nonlinearity. Output layer: single node output area quality score , mapped to the score interval by the Sigmoid function; Training method: using historical quality data (known level labels), using the Adam optimizer to minimize the mean square error (MSE) loss function.

[0090] Through fuzzy logic, the degree of thermal imaging matching, resistance anomaly amplitude, and confidence score are mapped to fuzzy levels of "mild / medium / severe", then input into the neural network to output the regional quality score, converting multi-dimensional indicators into a single quantitative score to achieve an intuitive assessment of regional quality; the score takes into account the influence of multiple factors and more objectively reflects the actual quality level of the region; the multi-dimensional indicators are converted into a single quantitative score to achieve an intuitive assessment of regional quality; the score takes into account the influence of multiple factors and more objectively reflects the actual quality level of the region.

[0091] Step S6: combining the defect classification result with the regional quality score to generate a detection region quality grade label, and implementing zoning detection of board quality based on the regional quality grade label.

[0092] The defect classification results are combined with the regional quality score to generate the quality grade label of the inspection area: The quality level label is determined by a decision tree, specifically: Construct a CART decision tree model based on regional quality score ( ) and defect type ( ) as the core feature and output quality level label ( ).

[0093] The decision tree is constructed as follows: Split conditions: Priority is given to the following areas based on their quality rating: Stable area; Further determine the defect type; Areas of potential defects or failures.

[0094] The second level is refined according to the defect type: If the defect type If it is a "fatal defect" (such as power short circuit), it will be directly determined as a fault area; if it is a "repairable defect" (such as a solder joint), the level will be adjusted according to the score (for example, a score of 65 points Slightly abnormal area).

[0095] Pruning optimization: avoid overfitting through cross-validation, set the minimum number of samples to 10 and the maximum depth to 5.

[0096] Example decision path: enter: , "Weak solder joint" → The decision tree determines it as a slightly abnormal area; input: , "Power layer short circuit" → The decision tree determines it as the fault area.

[0097] Detailed quality inspection report generation: Report structure design Detection overview: detection area coordinates, time-consuming, device model; Key indicator summary: thermal imaging matching degree, resistance anomaly amplitude, quality level.

[0098] Multi-dimensional analysis: thermal imaging: temperature cloud comparison (actual measurement vs. standard), thermal gradient distribution thermograph; resistance scanning: three-dimensional graph of defect cluster area resistance distribution, anomaly amplitude statistics; Defect diagnosis: matched historical sample comparison, confidence score feature contribution analysis; Decision basis: the calculation process of the area quality score is: (fuzzy rule + neural network weight visualization); decision tree judgment path diagram (showing the logic branch from input to level label); Improvement suggestion: repair scheme: repair steps and tool recommendations for defect types; Process optimization: correlate historical data and propose parameter adjustment strategies (e.g. weld temperature needs to be increased by 10℃). Realize the intelligent upgrade from data acquisition to quality decision, provide scientific, transparent, traceable quality evaluation system for board production.

[0099] Based on the defect classification results (such as fatal defects, repairable defects) and the area quality score, the quality level label (such as stable area, failure area) is generated through the CART decision tree, and finally the board quality partition detection is realized.

[0100] The present application can intuitively reflect the thermal field distribution of each area of the board through thermal imaging technology, and potential thermal abnormal areas can be found by comparing with the standard template. The resistance scanning can detect the conductivity of the board, and the defect cluster area can be determined by analyzing the resistance data and its gradient change. And by using data processing and machine learning algorithm, the collected data can be comprehensively analyzed to realize the accurate evaluation of the board quality.

[0101] It should be noted that in this text, relationship terms such as first and second are only used to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between the entities or operations. Moreover, the term "includes", "contains" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or equipment. Without more limitation. The statement "includes a limited element" does not exclude the existence of other identical elements in the process, method, article or equipment including the element.

[0102] While embodiments of the application have been shown and described, it is to be understood that the embodiments described are merely exemplary of the principles and application of the present application. Numerous modifications and adaptions can be effected without departing from the spirit and scope of the present application, which is not limited to the exact construction and arrangement described.

Claims

1. A method for detecting the quality of a board by partition, characterized in that: The following steps are involved: Step S1: Divide the board to be tested into regions according to its functional structure to obtain functional blocks of the board; perform thermal imaging acquisition on the functional blocks of the board to obtain thermal field distribution characteristic data; Step S2: Constructing standard thermal template data, using the z-score maximum method to quantify the difference between the thermal field distribution characteristic data of each board functional block and the standard thermal template data, and calculating the thermal imaging matching degree of each functional block. When the thermal imaging matching degree of a board functional block is greater than a preset threshold, the board functional block is marked as an abnormal thermal response area; Step S3: collecting resistance data and resistance gradient change data of the abnormal thermal response area. When the resistance value data or resistance gradient change data meets the preset defect cluster area locking condition, the abnormal thermal response area is further determined to be a defect cluster area. At the same time, based on the resistance value and resistance gradient change data of the abnormal thermal response area, the resistance abnormality amplitude is calculated; Step S4: combining the thermal field distribution characteristic data, resistance data and resistance gradient change data of the defect cluster area to obtain a composite characteristic map; Constructing a defect sample matching model, inputting the composite feature map into the defect sample matching model, and outputting a defect classification result and a confidence score; Step S5: According to a fuzzy logic reasoning and neural network combined algorithm, the thermal imaging matching degree, the resistance anomaly amplitude and the confidence score are combined to calculate the regional quality score; Step S6: combining the defect classification result with the regional quality score to generate a detection region quality grade label, and implementing zoning detection of board quality based on the regional quality grade label.

2. A method for detecting board quality by partition according to claim 1, characterized in that: The step S1 comprises: Based on the core principles of functional independence and signal flow of the board, combined with the modular design in the circuit diagram, the circuit units that perform the same function are divided into a block, including: power module, main control module, interface module and power load module; Each block corresponds to a thermal imaging unit that can be analyzed independently. The infrared thermal imager performs multi-state scanning on the powered board to collect temperature data of each functional block. Based on the temperature data of each functional block, the thermal field distribution cloud map, temperature rise curve and hot spot morphological characteristics are generated.

3. A method for detecting board quality by partition according to claim 2, characterized in that: The construction of standard thermal template data includes: Standard thermal template data construction: Collect thermal imaging data of qualified boards under the same working conditions, remove outliers and calculate the statistical feature data of each block: Temperature mean = the average temperature of the block of all qualified boards; Temperature standard deviation = reflects the temperature fluctuation range between qualified boards; Hot spot morphological characteristics: including the area, shape factor, and centroid position of a normal hot spot; Template update mechanism: After every 500 boards are inspected, new qualified data is automatically incorporated to recalculate μ and σ to adapt to process fluctuations.

4. A method for detecting board quality by partition according to claim 3, characterized in that: The z-score maximum value method is used to quantify the difference between the thermal field distribution characteristic data of each functional block and the standard thermal template data, and calculate the thermal imaging matching degree of each functional block, including: The difference between the measured data and the standard thermal template data is quantified by the z-score maximum method to obtain the thermal imaging matching degree. The formula is as follows: Calculate the z-score of each point: ； in, It is The temperature measurement value of each point; It is The mean temperature of each point; It is The standard deviation of the temperature at each point; Among all the calculated z-scores, take the maximum value : ; Define the degree of match: ； in, is the degree of thermal imaging matching, k is the attenuation coefficient, which determines the attenuation speed, is the point in the data set that deviates most from the mean. Range ∈(0,1], The higher the value, the closer the measured thermal field is to the standard state. ≥0.9: good match, no significant thermal anomaly; 0.7< <0.9: Slight deviation, needs to be analyzed in combination with other dimensions; ≤0.7: Significant abnormality, marked as "abnormal thermal response area".

5. A method for detecting board quality by partition according to claim 4, characterized in that: The collecting of resistance data and resistance gradient change data of the abnormal thermal response area includes: Contact probe: Suitable for high-precision detection. The probe tip directly contacts the surface of the object being measured to accurately collect surface resistance or electrical signal changes. Precision mechanical contact ensures detection stability and reliability. Meshing: The abnormal thermal response area is divided into two-dimensional grids with a spacing of 0.5 mm, that is, the area of ​​10 mm × 10 mm is divided into 20 × 20 = 400 grid points; Each grid point is marked as (i, j), corresponding to the board coordinates ( , ); Point-by-point scanning process: The probe positioning servo motor drives the probe array to move to just above point (i, j), and the contact probe applies a pressure of 10 g; For resistance measurement, apply a 100mV excitation voltage, measure the current flowing through the probe, and calculate the surface resistance using Ohm's law: ； in, is the resistance value at the (i, j) grid point, V is the input voltage, and I is the resistance value; Gradient calculation of the resistance gradient between the current point and the four adjacent points: ； in, Indicates the current point resistance gradient; Represents the resistance value of the current point; 、 Respectively represent the resistance values ​​of the adjacent points above and below the current point; 、 They represent the resistance values ​​of the adjacent points to the right and left of the current point respectively. The gradient value reflects the degree of mutation in the resistance distribution. The larger the gradient, the higher the possibility of defects.

6. A method for detecting board quality by partition according to claim 5, characterized in that: The calculation of the resistance abnormality amplitude based on the resistance value and resistance gradient change data of the abnormal thermal response area includes: The regional resistance anomaly amplitude is calculated by using the RMS weighted deviation method. The formula is as follows: ； in, is the resistance anomaly amplitude, It is the first The resistance value of each abnormal grid point; is the average normal resistance of the area; is the weight coefficient, which is set inversely proportional to the distance between the grid point and the cluster center; is the total number of abnormal grid points; the physical meaning of the numerator is the weighted square sum of the differences between the resistance values ​​of each abnormal grid point in the defect cluster area and the average normal resistance of the area, and the physical meaning of the denominator is the sum of the weight coefficients of all abnormal grid points.

7. A method for detecting board quality by partition according to claim 6, characterized in that: The composite characteristic map is obtained by combining the thermal field distribution characteristic data, resistance data and resistance gradient change data of the defect cluster area, including: The following key features are extracted from the defect cluster area to construct a composite feature map: Among them, thermal imaging features include: Hot spot centroid coordinates: ( , ), reflecting the location of the hot spot on the board; the maximum value of the thermal gradient: , indicating the most intense degree of thermal field change in the region; Average temperature: , reflecting the overall fever level of the area; Resistance sweep features include: Average resistance value: , which measures the average level of electrical conductivity in the area; Resistance gradient mean: , describes the average degree of change in the spatial distribution of resistance; Impedance reversal times: , record the number of "high-low-high" or "low-high-low" resistance fluctuations in the area; Defect cluster area: , represents the area of ​​abnormal resistance region; Integrate the above features into a composite feature map : 。 8. A method for detecting board quality by partition according to claim 7, characterized in that: The output is a defect classification result and confidence score, including: Use the cosine similarity algorithm to calculate the feature vector of the sample to be tested and historical sample feature vector Similarity : ； in, is the similarity between the detection sample feature and the historical sample feature vector, is the detection sample feature vector, is the historical sample feature vector, The value range is , the closer the value is to , which means the more similar the two vectors are, the numerator is This is the calculation formula for the dot product of two vectors, which means the sum of the products of the corresponding elements in the feature vectors of the sample to be tested and the historical sample. is the jth feature vector in the detection sample feature vector, is the jth eigenvector in the historical sample eigenvector; Defect type output and confidence score: Select the one with the highest similarity Historical samples are sorted from high to low according to similarity, and the predicted defect type is the defect type corresponding to the sample with the highest similarity. , confidence score By the highest similarity Convert to: ； in, is the confidence score, is the highest similarity.

9. A method for detecting board quality by partition according to claim 8, characterized in that: The fuzzy logic reasoning combined with the neural network algorithm combines the thermal imaging matching degree, the resistance anomaly amplitude and the confidence score to calculate the regional quality score, including: Nonlinear construction of regional quality scoring function: By combining fuzzy logic reasoning with neural networks, the thermal imaging matching degree, resistance anomaly amplitude, and confidence score are converted into regional quality scores. Among them, the fuzzy logic preprocessing is: Define fuzzy sets: map raw data to "low / medium / high" fuzzy levels: The thermal imaging matching degree is: Low: ,middle: ,high: ; The abnormal resistance amplitude is: Low ),middle ,high: ; The confidence scores are: Low: ,middle: ,high: ; Among them, the neural network includes: Input layer: takes the fuzzified features × 3 dimensions and the original confidence scores as input; Hidden layer: uses 2 fully connected layers and ReLU as the activation function; The first fully connected layer includes: Construct the weight matrix: ; Dimension meaning: number of input features × number of neurons Parameter function: each weight Indicates the The input feature pair The influence strength of each neuron; The calculation process is: =ReLU( · + ); in, is the first layer output vector, the feature representation after ReLU activation, Y is the 10-dimensional input vector, including the fuzzy features and the original confidence, is a 16-dimensional bias vector used to adjust the neuron activation threshold. ReLU is the activation function: nonlinearity is introduced, and the formula is ReLU( )=max(0, ), filter negative signals; The second fully connected layer includes: Weight matrix: ; Dimension meaning: 16 first layer output number × 8 second layer neuron number Parameter function: compress the high-dimensional features extracted by the first layer into a more abstract representation The calculation process is: ; in, The second layer outputs the vector, the final abstract feature representation, is the 16-dimensional vector output by the first layer, is an 8-dimensional bias vector, It is an activation function used to introduce nonlinear characteristics; Output layer: Single node outputs regional quality score , mapped to the scoring interval through the Sigmoid function; Training method: Use historical quality data and adopt Adam optimizer to minimize the mean square error loss function.

10. A method for detecting board quality by partition according to claim 9, characterized in that: Combining the defect classification result with the regional quality score to generate a quality grade label for the detection area includes: The quality level label is determined by a decision tree, specifically: Construct a CART decision tree model to score regional quality and defect types As the core feature, output quality level label ; The decision tree is constructed as follows: Split conditions: Priority is given to the following areas based on their quality rating: Stable area; Further determine the defect type; Potential defect or failure areas; The second level is refined according to the defect type: If the defect type If it is a "fatal defect", it will be directly determined as a fault area; if it is a "repairable defect", the level will be adjusted based on the score; Pruning optimization: avoid overfitting through cross-validation, set the minimum number of samples to 10 and the maximum depth to 5; Decision Path: enter: , "Weak solder joint" → The decision tree determines it as a slightly abnormal area; input: , "Power layer short circuit" → The decision tree determines it as the fault area.

Images