Al-based integrated circuit test management system
Through the AI-based integrated circuit test management system, deep learning technology is used to analyze the test data of integrated circuits, the problems of traditional low detection efficiency and insufficient accuracy are solved, and faster and more accurate fault detection is achieved.
Patent Information
- Application Number
- CN202410927494.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-11
- Publication Date
- 2024-11-08
AI Technical Summary
The traditional manual detection process is time-consuming and inefficient, and is not suitable for the needs of large-scale production and rapid detection. As the complexity of integrated circuit design increases, traditional fault judgment methods are difficult to adapt to new testing needs, affecting the accuracy of the results.
Using an AI-based integrated circuit test management system, the time series of test data of integrated circuits, including voltage, current and frequency, uses deep learning technology to perform timing feature analysis and interrelated interaction, and intelligently determine whether there are potential faults in the test point and bad.
Reliance on manual analysis is reduced, response speed is accelerated, abnormal patterns in test data can be more accurately identified, and fault detection is improved accuracy and timely.
Smart Images

Figure CN118916670A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of intelligent management, and more specifically, to an AI-based integrated circuit test management system. Background Art
[0002] Integrated circuit test management is a series of processes to ensure that integrated circuits (ICs) meet predetermined standards during design and production, which involves multiple testing stages, including design verification, wafer testing, post-packaging testing, etc. These tests not only involve electrical performance, but also include the evaluation of the physical and chemical properties of the chip. Effective test management can improve production efficiency, reduce costs, and ensure product quality.
[0003] In integrated circuits, potential failures or defects may cause the device to not work properly or output errors, affecting the performance and reliability of electronic equipment. For example, power supply problems may prevent the integrated circuit from starting or running; overheating may damage internal components; and connection problems may cause signal transmission interruption. Therefore, it is crucial to thoroughly test the integrated circuit to ensure quality and reliability.
[0004] However, the traditional manual inspection process is time-consuming and inefficient, which is not suitable for large-scale production and rapid inspection, resulting in low efficiency. Secondly, as the complexity of integrated circuit design increases, traditional fault judgment methods are difficult to adapt to new testing requirements, affecting the accuracy of the results.
[0005] Therefore, an integrated circuit test management system based on Al is expected. Summary of the invention
[0006] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides an integrated circuit test management system based on Al, which collects the time series of integrated circuit test data, wherein the test data includes voltage, current and frequency, and uses Al and deep learning technology to perform timing feature analysis and mutual correlation interaction on the test data, so as to intelligently obtain the judgment result of whether the current test point has potential faults and defects. In this way, the dependence on manual analysis is reduced, while the response speed is accelerated, and abnormal patterns in test data can be more accurately identified, so as to timely discover potential faults and defects, and improve the accuracy and timeliness of fault detection.
[0007] According to one aspect of the present application, there is provided an integrated circuit test management system based on AI, comprising:
[0008] A test data acquisition module, used to collect a time series of test data of the integrated circuit, wherein the test data includes voltage, current and frequency;
[0009] A test data regularization module, used for regularizing the time series of the test data according to the time dimension and the sample dimension to obtain a sequence of test data time series input vectors;
[0010] A test data local short-term temporal coding module, used for performing test data local short-term temporal coding on the sequence of the test data temporal input vectors to obtain a sequence of test data short-term temporal pattern feature vectors;
[0011] A test data long-term temporal sequence encoding module, used for performing test data long-term temporal sequence encoding on the sequence of the test data temporal sequence input vectors to obtain a sequence of test data long-term temporal sequence pattern feature vectors;
[0012] A feature fusion module, used for fusing the sequence of the short-term time series pattern feature vectors of the test data and the sequence of the long-term time series pattern feature vectors of the test data to obtain a sequence of multi-scale time series pattern feature vectors of the test data;
[0013] A test data feature association interaction module, used for performing test data feature association interaction on the sequence of the test data multi-scale time series pattern feature vectors to obtain a test data feature association interaction feature matrix;
[0014] The judgment result generating module is used to obtain the judgment result based on the test data feature association interaction feature matrix.
[0015] Compared with the prior art, the present application provides an integrated circuit test management system based on Al, which collects the time series of integrated circuit test data, wherein the test data includes voltage, current and frequency, and uses Al and deep learning technology to perform timing feature analysis and mutual correlation interaction on the test data, so as to intelligently obtain the judgment result of whether the current test point has potential faults and defects. In this way, the dependence on manual analysis is reduced, while the response speed is accelerated, and abnormal patterns in test data can be more accurately identified, so as to timely discover potential faults and defects, and improve the accuracy and timeliness of fault detection. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.
[0017] Figure 1 4 is a block diagram of an integrated circuit test management system based on Al according to an embodiment of the present application.
[0018] Figure 2 Schematic diagram of the architecture of an integrated circuit test management system based on Al according to an embodiment of the present application.
[0019] Figure 3 4 is a block diagram of a judgment result generation module in an integrated circuit test management system based on Al according to an embodiment of the present application.
[0020] Figure 4 4 is a block diagram of a feature optimization unit in an Al-based integrated circuit test management system according to an embodiment of the present application. DETAILED DESCRIPTION
[0021] The embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments described herein, but rather these embodiments are provided for a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.
[0022] It should be understood that the various steps described in the method embodiments of the present disclosure may be performed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present disclosure is not limited in this respect.
[0023] It should be noted that the terms "first\second\third" involved in the embodiments of the present application are only used to distinguish similar objects, and do not represent a specific order for the objects. It is understandable that the specific order or sequence of "first\second\third" can be interchanged where permitted. It should be understood that the objects distinguished by "first\second\third" can be interchanged where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein.
[0024] As shown in the present disclosure and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.
[0025] Various exemplary embodiments, features and aspects of the present disclosure will be described in detail below with reference to the accompanying drawings. The same reference numerals in the accompanying drawings represent elements with the same or similar functions. Although various aspects of the embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless otherwise specified.
[0026] Integrated circuit test management is a series of processes to ensure that integrated circuits (ICs) meet predetermined standards during design and production, which involves multiple testing stages, including design verification, wafer testing, post-packaging testing, etc. These tests not only involve electrical performance, but also include the evaluation of the physical and chemical properties of the chip. Effective test management can improve production efficiency, reduce costs, and ensure product quality.
[0027] In integrated circuits, potential failures or defects may cause the device to not work properly or output errors, affecting the performance and reliability of electronic equipment. For example, power supply problems may prevent the integrated circuit from starting or running; overheating may damage internal components; and connection problems may cause signal transmission interruptions. Therefore, it is critical to thoroughly test integrated circuits to ensure quality and reliability.
[0028] However, the traditional manual inspection process is time-consuming and inefficient, which is not suitable for large-scale production and rapid inspection, resulting in low efficiency. Secondly, as the complexity of integrated circuit design increases, traditional fault judgment methods are difficult to adapt to new testing requirements, affecting the accuracy of the results.
[0029] It should be understood that AI usually refers to Artificial Intelligence, which is a branch of computer science that aims to create systems that can perform tasks that usually require human intelligence. Artificial intelligence technologies include machine learning, natural language processing, computer vision, and expert systems. With the continuous development of AI, the application of AI in integrated circuit test management is becoming more and more important. AI can improve the quality and speed of integrated circuit testing by automating test processes, optimizing test strategies, and improving the analysis efficiency of test data. For example, AI algorithms can predict defects that may occur in chips during the production process, thereby eliminating potential problems at an early stage. In addition, AI is also able to process large amounts of test data and identify patterns and trends through machine learning models, which is crucial to improving the yield and reliability of chips. With the rapid development of the integrated circuit industry, the application of AI in test management is also constantly improving. It can perform complex test sequences, reduce human errors, and improve the consistency and accuracy of tests.
[0030] Therefore, an AI-based integrated circuit test management system is expected, which collects the time series of integrated circuit test data, wherein the test data includes voltage, current and frequency, and uses AI and deep learning technology to perform time series feature analysis and mutual correlation interaction on the test data, so as to intelligently obtain the judgment result of whether the current test point has potential faults and defects. In this way, the dependence on manual analysis is reduced, the response speed is accelerated, and abnormal patterns in test data can be more accurately identified, so as to timely discover potential faults and defects, and improve the accuracy and timeliness of fault detection.
[0031] Figure 1 4 is a block diagram of an integrated circuit test management system based on Al according to an embodiment of the present application. Figure 2 FIG. 1 is a schematic diagram of the architecture of an integrated circuit test management system based on Al according to an embodiment of the present application. Figure 1 and Figure 2 As shown, according to an embodiment of the present application, an integrated circuit test management system 100 based on Al includes: a test data acquisition module 110, which is used to collect the time series of the test data of the integrated circuit, and the test data includes voltage, current and frequency; a test data data regularization module 120, which is used to regularize the time series of the test data according to the time dimension and the sample dimension to obtain a sequence of test data timing input vectors; a test data local short-term timing encoding module 130, which is used to perform test data local short-term timing encoding on the sequence of the test data timing input vectors to obtain a sequence of test data short-term timing pattern feature vectors; a test data long-term timing encoding module 140, which is used to regularize the time series of the test data according to the time dimension and the sample dimension to obtain a sequence of test data timing input vectors; The sequence of test data time series input vectors is subjected to test data long-term time series encoding to obtain a sequence of test data long-term time series pattern feature vectors; a feature fusion module 150 is used to fuse the sequence of test data short-term time series pattern feature vectors and the sequence of test data long-term time series pattern feature vectors to obtain a sequence of test data multi-scale time series pattern feature vectors; a test data feature association interaction module 160 is used to perform test data feature association interaction on the sequence of test data multi-scale time series pattern feature vectors to obtain a test data feature association interaction feature matrix; and a judgment result generation module 170 is used to obtain a judgment result based on the test data feature association interaction feature matrix.
[0032] In an embodiment of the present application, the test data acquisition module 110 is used to collect the time series of the test data of the integrated circuit, and the test data includes voltage, current and frequency. It should be understood that considering that the voltage, the current and the frequency are the basic working parameters of the electronic circuit, they play a vital role in the test of the integrated circuit (IC). In detail, the voltage is a physical quantity that measures the potential difference between two points in the circuit, usually in volts (V). In the integrated circuit test, the voltage parameter is used to ensure that the power supply of the circuit meets the design specifications and monitor the power supply stability of the circuit. The current is the charge flow through the cross section of the conductor per unit time, usually in amperes (A). The current measurement in the test helps to detect abnormal flow in the circuit, such as short circuit or overload. The frequency is the number of periodic events that occur per unit time, usually in Hertz (Hz). In the integrated circuit test, the frequency parameter is usually used to evaluate the performance of the clock signal or oscillator to ensure that they work at a predetermined rate. Based on this, in order to accurately evaluate whether the test points of the integrated circuit are potentially faulty and defective, in the technical solution of the present application, the time series of the integrated circuit's test data is collected, and the test data includes voltage, current and frequency, and timing feature analysis and timing correlation interaction are performed on them. This can promptly indicate faults or defects in the circuit, help ensure product quality, avoid damage or safety hazards, which is of great significance for improving product reliability, reducing return rates and maintenance costs.
[0033] In an embodiment of the present application, the test data data regularization module 120 is used to regularize the time series of the test data according to the time dimension and the sample dimension to obtain a sequence of test data time series input vectors. Accordingly, considering that the time series of the test data is collected in the order of time, and the test data including voltage, current and frequency are constantly changing with the change of time, and there is a mutual correlation relationship between the voltage, current and frequency in time. Based on this, in the technical solution of the present application, the time series of the test data is regularized according to the time dimension and the sample dimension to obtain a sequence of test data time series input vectors. In other words, the regularization of the time dimension enables the time series of the test data to maintain the time order, which is very important for time series analysis, and can reveal the time series trends, periodic changes and seasonal characteristics in the test data. The regularization of the sample dimension can ensure that all relevant test data at each time point are available, and each data set in the test data can be analyzed and compared separately, which is helpful for cross-sample time series comparison and analysis.
[0034] In an embodiment of the present application, the test data local short-term temporal encoding module 130 is used to perform test data local short-term temporal encoding on the sequence of the test data temporal input vectors to obtain a sequence of test data short-term temporal pattern feature vectors. Specifically, in an embodiment of the present application, the test data local short-term temporal encoding module is used to: input the sequence of the test data temporal input vectors into a test data local short-term temporal encoding network based on 1D-CNN to obtain a sequence of test data short-term temporal pattern feature vectors. It should be understood that, considering that each test data temporal input vector in the sequence of the test data temporal input vectors has temporal feature information about the test data in a local time period. Therefore, in order to capture and mine the local temporal characteristics of each test data in the sequence of the test data temporal input vectors, in the technical solution of the present application, the sequence of the test data temporal input vectors is input into a test data local short-term temporal encoding network based on 1D-CNN to obtain a sequence of test data short-term temporal pattern feature vectors. It is worth mentioning that the 1D-CNN is particularly good at extracting local time series features from one-dimensional sequence data, which is very important for understanding short-term patterns in test data. In addition, the one-dimensional convolution operation can capture local time series dependencies in time series data, which helps to identify short-term dynamic changes in the data. That is, the local short-term time series encoding network for test data based on 1D-CNN can capture the local time series features of the test data in each test data time series input vector, such as instantaneous changes or periodic fluctuations in the short term, etc., and can effectively extract short-term time series pattern features in the test data, providing strong data support for subsequent analysis and decision-making.
[0035] In an embodiment of the present application, the test data long-term temporal encoding module 140 is used to perform test data long-term temporal encoding on the sequence of the test data temporal input vectors to obtain a sequence of test data long-term temporal pattern feature vectors. Specifically, in an embodiment of the present application, the test data long-term temporal encoding module is used to: input the sequence of the test data temporal input vectors into a test data long-term temporal encoding network based on LSTM to obtain a sequence of the test data long-term temporal pattern feature vectors. Accordingly, considering that each test data temporal input vector in the sequence of the test data temporal input vectors also contains test data temporal features under long-distance time scale dependence. Therefore, in order to mine the temporal correlation relationship of each test data under a long time span from the sequence of the test data temporal input vectors, in the technical solution of the present application, the sequence of the test data temporal input vectors is input into a test data long-term temporal encoding network based on LSTM to obtain a sequence of the test data long-term temporal pattern feature vectors. It is worth mentioning that the LSTM is specially designed to solve the long-term dependency problem in traditional RNN (recurrent neural network). It can learn the long-term correlation in time series data. In particular, the LSTM has memory units and gating mechanisms that control the flow of information, which can determine which information should be retained, forgotten or updated, which makes it very suitable for processing and predicting long-term patterns in time series. Therefore, by inputting the sequence of the test data timing input vector into the LSTM-based test data long-term timing encoding network, the long-term timing pattern characteristics and change laws in the various test data timing input vectors can be effectively captured and learned, which is very valuable for understanding the behavior of the circuit, predicting potential failures and optimizing the test process.
[0036] In an embodiment of the present application, the feature fusion module 150 is used to fuse the sequence of the test data short-term time series pattern feature vectors and the sequence of the test data long-term time series pattern feature vectors to obtain a sequence of test data multi-scale time series pattern feature vectors. Specifically, in an embodiment of the present application, the feature fusion module is used to: fuse the sequence of the test data short-term time series pattern feature vectors and the sequence of the test data long-term time series pattern feature vectors by position to obtain a sequence of the test data multi-scale time series pattern feature vectors. It should be understood that each feature vector in the sequence of the test data short-term time series pattern feature vectors and the sequence of the test data long-term time series pattern feature vectors expresses the local short-term time series pattern feature information and the long-term pattern time series feature information of different test data. Therefore, in order to fuse the short-term time series pattern features of the test data and the long-term time series pattern features of the test data, so as to integrate the time series information of the test data at different time scales and provide a more comprehensive data analysis for the model, in the technical solution of the present application, the sequence of the short-term time series pattern feature vectors of the test data and the sequence of the long-term time series pattern feature vectors of the test data are fused by position to obtain a sequence of multi-scale time series pattern feature vectors of the test data. Specifically, the short-term time series pattern features of the test data capture the rapid changes or instantaneous events of the test data, while the long-term time series pattern features of the test data reflect the chronic changes or trends of the test data. By fusing these features, they can complement each other and capture the local details and global trends of each test data at the same time, thereby obtaining a more comprehensive and integrated multi-scale time series pattern feature information of the test data, which provides help and support for the fault judgment of subsequent test points.
[0037] In an embodiment of the present application, the test data feature association interaction module 160 is used to perform test data feature association interaction on the sequence of the test data multi-scale time series pattern feature vectors to obtain a test data feature association interaction feature matrix. Specifically, in an embodiment of the present application, the test data feature association interaction module is used to: input the sequence of the test data multi-scale time series pattern feature vectors into a test data feature association interaction network based on a Gaussian feature fusion module to obtain the test data feature association interaction feature matrix. In particular, in a specific embodiment of the present application, the Gaussian feature fusion module is a feature fusion module based on a Gaussian density map. Accordingly, considering that the sequence of the test data multi-scale time series pattern feature vectors respectively represents the multi-scale time series pattern feature information of different test data (voltage, current and frequency) in the short-term and long-term time series. Therefore, in order to integrate and aggregate the sequence of the test data multi-scale time series pattern feature vectors into a comprehensive test data interaction feature, in the technical solution of the present application, the sequence of the test data multi-scale time series pattern feature vectors is input into a test data feature association interaction network based on a Gaussian feature fusion module to obtain the test data feature association interaction feature matrix. That is to say, the Gaussian feature fusion module can effectively combine the timing features from different types of test data to generate a comprehensive feature representation, which helps to capture the complex timing correlation interaction patterns and interactions in the data. At the same time, the Gaussian feature fusion can weight the multi-scale timing pattern features of different test data by adjusting the parameters of the Gaussian distribution (such as the mean and variance), highlight the multi-scale timing pattern features of important test data and suppress noise, thereby providing richer and more powerful information for the final data analysis, fault detection or prediction tasks.
[0038] It is worth mentioning that the Gaussian feature fusion module is a feature fusion technology based on Gaussian density map, which is usually used in machine learning and signal processing. The purpose is to effectively combine feature information from different sources or different scales to obtain a more comprehensive and robust feature representation. In detail, first, the Gaussian density is estimated for each independent feature set to determine its probability distribution parameters, such as the mean (μ), variance (σ 2) or covariance matrix. Then, the original features are mapped into the feature space of Gaussian distribution, in which each feature is associated with a Gaussian distribution that describes the statistical properties of the feature. After that, a weight is assigned to each Gaussian distribution according to the importance or reliability of each feature. Finally, in the Gaussian feature space, the interaction between different features can be represented by the parameters of the Gaussian distribution (such as mean and covariance), so as to capture the correlation and interaction between features, and the fused features are reduced and transformed by linear or nonlinear methods to obtain feature representations that are more suitable for subsequent processing or analysis. Therefore, the advantage of the Gaussian feature fusion module is that it can handle uncertainty and variability, while being able to express the complex relationship between features through a probabilistic model. This module is particularly useful in dealing with scenarios with high-dimensional data and multimodal data, such as image processing, speech recognition, bioinformatics, and financial analysis. In integrated circuit test management, the Gaussian feature fusion module can help extract more accurate and robust features from test data to improve the accuracy of fault detection and prediction.
[0039] In the embodiment of the present application, the judgment result generating module 170 is used to obtain a judgment result based on the test data feature association interaction feature matrix. Figure 3 FIG. 1 is a block diagram of a judgment result generation module in an integrated circuit test management system based on A1 according to an embodiment of the present application. Specifically, in the embodiment of the present application, Figure 3 As shown, the judgment result generation module 170 includes: a feature optimization unit 171, used to perform feature optimization on the test data feature association interaction feature matrix to obtain an optimized test data feature association interaction feature matrix; and a test point defect assessment unit 172, used to input the optimized test data feature association interaction feature matrix into a classifier-based fault determiner to obtain a judgment result on whether the current test point has potential faults and defects.
[0040] Specifically, the feature optimization unit 171 is used to perform feature optimization on the test data feature association interaction feature matrix to obtain an optimized test data feature association interaction feature matrix. Specifically, in an embodiment of the present application, the feature optimization unit is used to perform a deep correlation correction based on a coherent attention response mechanism on the test data feature association interaction feature matrix to obtain the optimized test data feature association interaction feature matrix. In particular, in the technical solution of the present application, the test data feature association interaction feature matrix is a key information carrier that integrates short-term and long-term temporal pattern features for final fault judgment. Therefore, improving the feature significance of the feature matrix and avoiding feature redundancy are crucial to ensuring the accuracy and efficiency of the fault detection system. Specifically, significant features can more accurately capture the patterns and trends related to faults in the test data and improve the accuracy of fault detection. Significant features help distinguish between normal states and potential fault states, reducing the possibility of misjudgment and missed judgment. In time series data, significant features can better adapt to changes in data and maintain high detection performance even when test conditions or environments change. At the same time, removing redundant features can reduce the computational burden of the model and improve the speed and efficiency of fault detection. Redundant features may cause the model to overfit, affecting its performance on unknown data. Avoiding feature redundancy helps improve the generalization ability of the model. Removing redundant features helps to simplify the model structure, making the model easier to understand and maintain. Significant features help to quickly identify faults, allowing the system to respond in a timely manner and reduce potential losses. By improving the significance of features, false positives and false negatives caused by irrelevant or minor information can be reduced, and the reliability of fault detection can be improved. Significant features and avoiding redundancy help improve the applicability of fault detection systems in different test scenarios and conditions. As the scale of the system expands and the amount of data increases, a system that efficiently and accurately processes features can be more easily expanded and maintained. Based on this, the test data feature association interaction feature matrix is subjected to deep correlation correction based on the coherence attention response mechanism to obtain an optimized test data feature association interaction feature matrix.
[0041] Figure 4 FIG. 1 is a block diagram of a feature optimization unit in an integrated circuit test management system based on Al according to an embodiment of the present application. Specifically, in the embodiment of the present application, Figure 4As shown, the feature optimization unit 171 includes: a feature flattening subunit 1711, used to perform feature flattening on the test data feature association interaction feature matrix to obtain a test data feature association interaction feature vector; a correlation correction subunit 1712, used to perform deep correlation correction on the test data feature association interaction feature vector based on a coherent attention response mechanism to obtain an optimized test data feature association interaction feature vector; and a reverse feature aggregation subunit 1713, used to perform reverse feature aggregation on the optimized test data feature association interaction feature vector in accordance with the feature flattening method to obtain the optimized test data feature association interaction feature matrix.
[0042] More specifically, in an embodiment of the present application, the correlation correction subunit is used to: determine the mean of the test data feature association interaction feature vector; accumulate and sum the test data feature association interaction feature vectors to obtain the sum value of the test data feature association interaction feature vectors; subtract the mean of the test data feature association interaction feature vector from the sum value of the test data feature association interaction feature vector, divide the result by a predetermined hyperparameter, and then divide the result by the length of the test data feature association interaction feature vector to obtain a first quotient value; calculate an exponential function value with a natural constant as the base for the first quotient value to obtain a first natural exponential function value; and perform point multiplication of the first natural exponential function value and the test data feature association interaction feature vector to obtain an optimized test data feature association interaction feature vector.
[0043] In the embodiment of the present application, specifically, the correlation correction subunit is used to: perform deep correlation correction based on the coherence attention response mechanism on the test data feature association interaction feature vector using the following feature correction formula to obtain an optimized test data feature association interaction feature vector, wherein; the feature correction formula is:
[0044]
[0045] Wherein, V1 represents the test data feature association interaction feature vector, v i represents the eigenvalue of the i-th position of the test data feature association interaction feature vector, represents the mean of the test data feature association interaction feature vector, L represents the length of the test data feature association interaction feature vector, ε represents a predetermined hyperparameter, ⊙ represents position point multiplication, and V1′ represents the optimized test data feature association interaction feature vector.
[0046] Specifically, in order to improve the feature significance of the test data feature association interaction feature matrix and avoid feature redundancy, in the technical solution of the present application, the test data feature association interaction feature matrix is subjected to a depth correlation correction based on a coherent attention response mechanism, which uses the global mean of the test data feature association interaction feature matrix as its dynamically formed deep anchor feature, and applies a coherent attention response mechanism based on the deep anchor feature to the original feature distribution, so as to perform a position scattering response deep correlation correction in a class divergence space with depth difference, thereby making the original feature distribution have a highly significant depth correlation distribution characteristic, so as to improve the feature significance of the test data feature association interaction feature matrix and avoid feature redundancy.
[0047] Specifically, the test point bad evaluation unit 172 is used to input the optimized test data feature association interaction feature matrix into the fault judgement device based on the classifier to obtain the judgment result of whether the current test point is a potential fault or bad. That is, the optimized test data feature association interaction feature matrix is used for classification processing to intelligently obtain the judgment result of whether the current test point is a potential fault or bad. In this way, the dependence on manual analysis can be reduced, the response speed can be accelerated, and the abnormal patterns in the test data can be more accurately identified, so as to facilitate the timely discovery of potential faults and badness, thereby improving the accuracy of fault detection.
[0048] In summary, the integrated circuit test management system 100 based on Al according to the embodiment of the present application is explained, which collects the time series of the test data of the integrated circuit, wherein the test data includes voltage, current and frequency, and uses Al and deep learning technology to perform timing feature analysis and mutual correlation interaction on the test data, so as to intelligently obtain the judgment result of whether the current test point has potential faults and defects. In this way, the dependence on manual analysis is reduced, while the response speed is accelerated, and abnormal patterns in the test data can be more accurately identified, so as to timely discover potential faults and defects, and improve the accuracy and timeliness of fault detection.
[0049] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from a website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (digital subscriber line, DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media integrated. The available medium may be a magnetic medium (eg, a floppy disk, a hard disk, a magnetic tape), an optical medium (eg, a digital video disc (DVD)), or a semiconductor medium (eg, a solid state disk (SSD)).
[0050] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.
[0051] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.
[0052] The above descriptions of various implementations of the present disclosure are exemplary and not exhaustive. The disclosure is not limited to the disclosed implementations, and many modifications and variations are obvious to those of ordinary skill in the art without departing from the scope and spirit of the described implementations. The terms used herein are selected to best explain the principles of the implementations, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the various practical methods disclosed herein.
Claims
1. An integrated circuit test management system based on Al, characterized in that: include: A test data acquisition module, used to collect a time series of test data of the integrated circuit, wherein the test data includes voltage, current and frequency; A test data regularization module, used for regularizing the time series of the test data according to the time dimension and the sample dimension to obtain a sequence of test data time series input vectors; A test data local short-term temporal coding module, used for performing test data local short-term temporal coding on the sequence of the test data temporal input vectors to obtain a sequence of test data short-term temporal pattern feature vectors; A test data long-term temporal sequence encoding module, used for performing test data long-term temporal sequence encoding on the sequence of the test data temporal sequence input vectors to obtain a sequence of test data long-term temporal sequence pattern feature vectors; A feature fusion module, used for fusing the sequence of the short-term time series pattern feature vectors of the test data and the sequence of the long-term time series pattern feature vectors of the test data to obtain a sequence of multi-scale time series pattern feature vectors of the test data; A test data feature association interaction module, used for performing test data feature association interaction on the sequence of the test data multi-scale time series pattern feature vectors to obtain a test data feature association interaction feature matrix; The judgment result generating module is used to obtain the judgment result based on the test data feature association interaction feature matrix.
2. The integrated circuit test management system based on Al according to claim 1, characterized in that: The test data local short-term temporal coding module is used for: The sequence of the test data time series input vectors is input into a test data local short-term time series encoding network based on 1D-CNN to obtain a sequence of the test data short-term time series pattern feature vectors.
3. The integrated circuit test management system based on Al according to claim 2, characterized in that: The test data long-term temporal coding module is used for: The sequence of the test data time series input vectors is input into the LSTM-based test data long-term time series encoding network to obtain a sequence of the test data long-term time series pattern feature vectors.
4. The integrated circuit test management system based on Al according to claim 3, characterized in that: The feature fusion module is used to: The sequence of the short-term temporal pattern feature vectors of the test data and the sequence of the long-term temporal pattern feature vectors of the test data are fused by position to obtain the sequence of the multi-scale temporal pattern feature vectors of the test data.
5. The integrated circuit test management system based on Al according to claim 4, characterized in that: The test data feature association interaction module is used to: The sequence of the test data multi-scale temporal pattern feature vectors is input into the test data feature association interaction network based on the Gaussian feature fusion module to obtain the test data feature association interaction feature matrix.
6. The integrated circuit test management system based on Al according to claim 5, characterized in that: The judgment result generating module comprises: A feature optimization unit, used for performing feature optimization on the test data feature association interaction feature matrix to obtain an optimized test data feature association interaction feature matrix; The test point badness evaluation unit is used to input the optimized test data feature association interaction feature matrix into a fault judgement device based on a classifier to obtain a judgment result on whether the current test point has a potential fault or badness.
7. The integrated circuit test management system based on Al according to claim 6, characterized in that: The feature optimization unit is used to: The test data feature association interaction feature matrix is subjected to deep correlation correction based on a coherent attention response mechanism to obtain the optimized test data feature association interaction feature matrix.
8. The integrated circuit test management system based on Al according to claim 7, characterized in that: The feature optimization unit comprises: A feature flattening subunit, used for flattening the test data feature association interaction feature matrix to obtain a test data feature association interaction feature vector; A correlation correction subunit, used for performing a deep correlation correction based on a coherent attention response mechanism on the test data feature correlation interaction feature vector to obtain an optimized test data feature correlation interaction feature vector; The reverse feature aggregation subunit is used to perform reverse feature aggregation on the optimized test data feature association interaction feature vector in the feature flattening manner to obtain the optimized test data feature association interaction feature matrix.
9. The integrated circuit test management system based on Al according to claim 8, characterized in that: The correlation correction subunit is used for: Determining a mean value of the test data feature association interaction feature vector; Accumulating and summing the test data feature-associated interactive feature vectors to obtain a sum value of the test data feature-associated interactive feature vectors; Subtracting the mean of the test data feature association interaction feature vector from the sum of the test data feature association interaction feature vectors and dividing the sum by a predetermined hyperparameter and then by the length of the test data feature association interaction feature vector to obtain a first quotient; Calculating an exponential function value with a natural constant as a base for the first quotient value to obtain a first natural exponential function value; The first natural exponential function value is point-multiplied by the test data feature association interaction feature vector to obtain an optimized test data feature association interaction feature vector.
Citation Information
Patent Citations
Self-test method for circuit fault of AI chip of intelligent soybean milk maker
CN108896905A
Failure prediction system of hydraulic system and method thereof
CN117231590A
Power transmission line fault identification system and method based on artificial intelligence
CN117368639A
Method and system for scanning test of integrated circuit
CN117805593A
Double-channel feature fusion bearing equipment fault diagnosis method based on attention mechanism
CN118193935A
Cited By
Wavelength division multiplexer chip post-processing management system, device and medium
CN120750418A
