CN-122023426-A - IC carrier plate defect detection method and system

CN122023426ACN 122023426 ACN122023426 ACN 122023426ACN-122023426-A

Abstract

The application provides a method and a system for detecting defects of an IC carrier plate, which relate to the technical field of defect detection of the IC carrier plate, and the method comprises the steps of uniformly dividing the IC carrier plate into a plurality of grid areas to be recorded as detection areas, and collecting detection data of each detection area in real time; extracting data characteristic parameters, calculating characteristic change rates of the data characteristic parameters, constructing a defect evolution sequence, constructing a defect evolution model, inputting the data characteristic parameters and the defect evolution sequence into the defect evolution model, outputting defect probability and corresponding defect types of each detection area, generating a defect evolution trend graph, and selecting corresponding decisions from a preset decision base according to the defect evolution trend graph to carry out optimization adjustment. According to the application, the defect evolution sequence is constructed by analyzing the characteristic parameters and the change rate of the detection region, so that the transition from static detection to trend prediction is realized, the early warning can be carried out before the defect is formed, the defect expansion risk is reduced, and the product quality stability is improved.

Inventors

WANG MENG
LIU DONGHU
MA JIFENG
LI ZHENLAI

Assignees

清河电子科技(山东)有限责任公司

Dates

Publication Date: 20260512
Application Date: 20260414

Claims (10)

1. A method for detecting defects of an IC carrier, the method comprising: The data acquisition, namely uniformly dividing the IC carrier plate into a plurality of grid areas to be recorded as detection areas, and acquiring detection data of each detection area in real time, wherein the detection data comprises electrical parameters, circuit images and structural data; Extracting data characteristic parameters based on detection data, wherein the data characteristic parameters comprise electrical characteristic parameters, image characteristic parameters and structural characteristic parameters, calculating characteristic change rates of the data characteristic parameters and constructing a defect evolution sequence; Constructing a model, namely constructing a defect evolution model based on a deep learning algorithm; Inputting the data characteristic parameters and the defect evolution sequence into a defect evolution model, outputting the defect probability of each detection area and the corresponding defect type, and generating a defect evolution trend graph; And (3) decision optimization, namely selecting a corresponding decision from a preset decision library to perform optimization adjustment based on the preset decision library according to the defect evolution trend graph.
2. The method for detecting defects of an IC carrier according to claim 1, wherein the step of extracting features specifically includes: The method comprises the steps of extracting instantaneous values, average values, variances, frequency domain features, waveform shape features, currents and impedances of voltages from electrical parameters, extracting gray histograms, texture features and edge densities from circuit images, extracting geometric dimensions, circuit structures and interlayer structures from structural feature parameters, calculating feature change rates of all data feature parameters, and arranging to obtain a defect evolution sequence.
3. The IC carrier defect detection method according to claim 1, further comprising, after the step of feature extraction, before the step of modeling: And performing multi-modal feature interaction, namely performing high-order tensor interaction on each data feature parameter to obtain interaction features, performing tensor decomposition on the interaction features to obtain high-order interaction features, and splicing each data feature parameter with the high-order interaction features to obtain multi-modal interaction features.
4. The IC carrier defect detection method of claim 3, further comprising, after the step of multi-modal feature interaction, before the step of modeling: And space aggregation, namely determining a neighborhood set of the current detection region based on a preset space adjacency relationship, acquiring multi-modal interaction characteristics of each detection region in the neighborhood set, and carrying out weighted aggregation on the multi-modal interaction characteristics of each detection region in the neighborhood set to acquire the space interaction characteristics of the current detection region.
5. The IC carrier defect detection method of claim 4, further comprising, after the step of spatially aggregating, before the step of modeling: And (3) time aggregation, namely acquiring multi-modal interaction characteristics of a plurality of time windows before the current time window of the current detection area and corresponding characteristic change rates based on a preset time window, and carrying out weighted fusion on the multi-modal interaction characteristics of the time windows to obtain the time interaction characteristics of the current detection area.
6. The IC carrier defect detection method of claim 5, further comprising, after the step of time-aggregating, before the step of modeling: And (3) feature fusion, namely carrying out weighted fusion on the multi-mode interaction features, the space interaction features and the time interaction features to obtain space-time interaction features, calculating the feature change rate of each space-time interaction feature and the feature change rate of each data feature parameter, and splicing, and updating the defect evolution sequence based on the splicing result to obtain a new defect evolution sequence.
7. The IC carrier defect detection method according to claim 1, further comprising, after the step of feature extraction, before the step of modeling: Acquiring a characteristic causal graph, namely acquiring historical data characteristic parameters and corresponding defect types of each detection area, taking the data characteristic parameters and the defect types of each detection area of an IC carrier plate as nodes, acquiring causal edges among the nodes by adopting a Granger causal inspection algorithm, and constructing a defect evolution causal graph based on the nodes and the corresponding causal edges; the step of constructing the model specifically comprises the step of constructing the defect evolution model based on a deep learning algorithm and a defect evolution causal graph.
8. The IC carrier defect detection method of claim 1, further comprising, after the step of modeling, before the step of defect prediction: The model training comprises the steps of obtaining historical detection data and corresponding defect types of each detection area, extracting historical data characteristic parameters, constructing a historical defect evolution sequence based on the historical data characteristic parameters, marking the historical defect evolution sequence and the corresponding defect types as a training set, defining a loss function, inputting the training set into a defect evolution model for iterative training, calculating model loss according to the loss function in the iterative training process, updating model gradients based on the model loss, carrying out iterative optimization on model parameters by adopting a gradient descent algorithm, obtaining a trained defect evolution model, and marking the trained defect evolution model as a new defect evolution model.
9. The method for detecting defects of an IC carrier according to claim 1, wherein the step of decision optimization comprises: Establishing a process and defect association knowledge base, tracing each process procedure and corresponding process parameters of a detection area according to the detection area where the defect is located, recording as defect process data, constructing a defect root cause analysis model, inputting the defect type and the corresponding defect process data into the defect root cause analysis model, outputting root cause probability distribution of each sequence of the defect, selecting corresponding decisions from the preset decision base according to the root cause probability distribution of each sequence of the defect based on the preset decision base, and performing optimization adjustment; the process and defect association knowledge base comprises mapping relations among each process procedure, corresponding process parameters, parameter drift modes and defect types.
10. An IC carrier defect detection system adapted for use in a method as claimed in any one of claims 1 to 9, the system comprising: The data acquisition module is used for uniformly dividing the IC carrier plate into a plurality of grid areas to be recorded as detection areas and acquiring detection data of each detection area in real time, wherein the detection data comprises electrical parameters, circuit images and structural data; The feature extraction module is used for extracting data feature parameters based on the detection data, including electrical feature parameters, image feature parameters and structural feature parameters, calculating feature change rates of the data feature parameters and constructing a defect evolution sequence; the model building module is used for building a defect evolution model based on a deep learning algorithm; the defect prediction module is used for inputting the data characteristic parameters and the defect evolution sequence into a defect evolution model, outputting the defect probability of each detection area and the corresponding defect type, and generating a defect evolution trend graph; The decision optimization module is used for selecting corresponding decisions from the preset decision library to perform optimization adjustment according to the defect evolution trend graph based on the preset decision library.

Description

IC carrier plate defect detection method and system Technical Field The application relates to the technical field of defect detection of an IC carrier plate, in particular to a method and a system for detecting the defect of the IC carrier plate. Background With the development of electronic information technology, the performance and reliability requirements of high-end chips on an IC carrier (integrated circuit carrier) are increasing. The IC carrier plate is used as an interface between the chip and an external circuit, has a complex structure, dense circuits and multi-layer stacking, and bears key functions such as electric transmission, heat dissipation, mechanical support and the like. In the fields of servers, communication, automobile electronics, high-performance computing and the like, the micro defects of the carrier plate can directly influence the performance of the chip and even cause the system failure. In the prior art, methods such as Automatic Optical Inspection (AOI), X-ray inspection (X-ray), and electrical testing are generally used to detect defects of an IC carrier board, so as to identify problems such as circuit breakage, voids or bubbles, and interface delamination. Wherein bubble defects are generally controlled by parameters such as position and diameter. However, the above method is mostly based on static detection, and only the formed defects can be identified, so that the potential defects which are not developed or are in the evolution process are difficult to discover in time. Meanwhile, the small-size bubbles are difficult to capture due to the limitation of detection resolution, but can be further expanded in the subsequent lamination, thermal cycle or electroplating process, so that the performance of the carrier plate is affected. Disclosure of Invention In order to solve the defects of the prior art, the application provides a method and a system for detecting defects of an IC carrier plate. In a first aspect, the present application provides a method for detecting defects of an IC carrier, which adopts the following technical scheme: a method of IC carrier defect detection, the method comprising: The data acquisition, namely uniformly dividing the IC carrier plate into a plurality of grid areas to be recorded as detection areas, and acquiring detection data of each detection area in real time, wherein the detection data comprises electrical parameters, circuit images and structural data; Extracting data characteristic parameters based on detection data, wherein the data characteristic parameters comprise electrical characteristic parameters, image characteristic parameters and structural characteristic parameters, calculating characteristic change rates of the data characteristic parameters and constructing a defect evolution sequence; Constructing a model, namely constructing a defect evolution model based on a deep learning algorithm; Inputting the data characteristic parameters and the defect evolution sequence into a defect evolution model, outputting the defect probability of each detection area and the corresponding defect type, and generating a defect evolution trend graph; And (3) decision optimization, namely selecting a corresponding decision from a preset decision library to perform optimization adjustment based on the preset decision library according to the defect evolution trend graph. Optionally, the step of extracting the features specifically includes: The method comprises the steps of extracting instantaneous values, average values, variances, frequency domain features, waveform shape features, currents and impedances of voltages from electrical parameters, extracting gray histograms, texture features and edge densities from circuit images, extracting geometric dimensions, circuit structures and interlayer structures from structural feature parameters, calculating feature change rates of all data feature parameters, and arranging to obtain a defect evolution sequence. Optionally, after the step of feature extraction, before the step of constructing the model, further comprising: And performing multi-modal feature interaction, namely performing high-order tensor interaction on each data feature parameter to obtain interaction features, performing tensor decomposition on the interaction features to obtain high-order interaction features, and splicing each data feature parameter with the high-order interaction features to obtain multi-modal interaction features. Optionally, after the step of multi-modal feature interaction, before the step of building the model, further comprising: And space aggregation, namely determining a neighborhood set of the current detection region based on a preset space adjacency relationship, acquiring multi-modal interaction characteristics of each detection region in the neighborhood set, and carrying out weighted aggregation on the multi-modal interaction characteristics of each detection region in the neighborhood set to