IC MultiSite automatic test system based on PXIe
By introducing data fusion, intelligent scheduling, and security testing units into the PXIe-based IC MultiSite automated testing system, and combining them with cloud platform collaboration, the problems of testing efficiency and data management have been solved, achieving efficient and accurate intelligent testing to meet the needs of semiconductor manufacturing.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-24
- Publication Date
- 2026-03-10
AI Technical Summary
Existing PXIe-based IC MultiSite automated testing systems suffer from problems such as limited testing efficiency due to the static nature of workstation resource allocation, imperfect error-proofing mechanisms, non-intelligent test data management, and lack of cloud platform collaboration, resulting in high false negative rates and insufficient exploitation of data value.
The system employs a data fusion unit for feature extraction and mapping, an intelligent scheduling unit for adaptive task splitting and fault prediction, an intelligent security testing unit for quantum encryption, and a cloud platform for deep collaboration to achieve real-time interaction between the testing system and remote data. It also integrates a nanoscale sensor array for high-precision measurement.
It improves test throughput, reduces single-chip test costs, and achieves improved test efficiency, accuracy, and intelligence, meeting the semiconductor manufacturing industry's needs for high throughput, high reliability diagnostics, and intelligent end-to-end processing.
Smart Images
Figure CN121633784A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the field of automation testing, in particular to an IC MultiSite automated test system based on PXIe. BACKGROUND
[0002] PXIe is an open test platform based on PCI Express bus, which is composed of a chassis, a backplane, a controller and peripheral modules. It inherits the modular design of PXI and the high-speed serial transmission capability of PCIe, supports bandwidth transmission up to several GB / s, far exceeding traditional buses. PXIe platform is widely used in wafer testing, post-packaging testing and failure analysis, and in the fields of industrial automation, aerospace and communication testing. PXIe supports bus protocol testing through modular design, integrates fault tree diagnosis, data playback and other intelligent analysis functions.
[0003] At present, the IC MultiSite automated test system based on PXIe still has significant deficiencies in technical implementation. First, the test efficiency is limited by the static nature of the workstation resource allocation. In the traditional scheme, the binding relationship between the instrument module and the test workstation is fixed, which leads to the fact that high-value instrument resources cannot be dynamically shared when multiple workstations are parallel, forming a bottleneck of resource pooling deficiency. In addition, the foolproof mechanism has the defect of intelligence. Most existing systems rely on preset threshold to judge abnormalities, lack real-time AI analysis of intermediate results of chips during testing, and are difficult to capture atypical abnormal states, resulting in an increase in missed detection rate. Furthermore, the test data management has a last mile fault phenomenon. The local system only completes basic data collection and execution, lacks deep cooperation with the cloud platform, and leads to the fact that complex data analysis and real-time visual report generation still need manual intervention, which cannot meet the needs of semiconductor manufacturing for intelligent processing of test data in the whole link. These deficiencies directly restrict the potential release of MultiSite testing in throughput, reliability diagnosis and data value mining. SUMMARY
[0004] The purpose of the present application is to provide an IC MultiSite automated test system based on PXIe, which solves the problem of abnormal capture caused by imperfect foolproof mechanism, improves test throughput and reduces single-chip test cost. At the same time, through deep cooperation with the cloud platform, real-time interaction between the test system and remote data is realized, and the test efficiency, accuracy, flexibility, intelligence and data value are comprehensively improved.
[0005] To achieve the above purpose, the present application provides the following technical scheme: The present application provides an IC MultiSite automated test system based on PXIe, which includes a data fusion unit, an intelligent scheduling unit, an adaptive test unit and an intelligent safety test unit, wherein: The data fusion unit: acquires original data packets for feature extraction and mapping, constructs a cross-domain association model using the four types of feature vectors after feature extraction, obtains three types of data, and obtains a fusion feature vector through dynamic weight distribution of the three types of data; The intelligent scheduling unit: performs adaptive task splitting on the fusion feature vector through a load balancing algorithm, and performs fault prediction and avoidance on the test records and data; The adaptive test unit: dynamically adjusts the test data and the fusion feature vector, and generates adaptive test data based on the fusion feature vector; The intelligent safety test unit: performs quantum encryption driving on the adaptive test data output by the adaptive test unit, and introduces a nanoscale sensor array to obtain nanoscale fault positioning information.
[0006] The data fusion unit includes a feature extraction module, a model construction module, and a weight distribution module, wherein: The feature extraction module: acquires four types of data, namely voltage range, current tolerance, clock jitter, ADC signal-to-noise ratio, and high-temperature life, as original data packets, and features electrical specifications, timing parameters, functional specifications, and reliability data to obtain four types of feature vectors, namely dynamic range, synchronization accuracy, performance indicators, and life prediction; The model construction module: uses the four types of feature vectors after feature extraction to establish a graph neural network association model, uses a multi-modal fusion algorithm on the association model, quantizes the association strength of the association model, and obtains a cross-domain association graph and a global feature association matrix; The weight distribution module: performs real-time weight calculation on real-time test data, uses an adaptive weighting algorithm, combines a historical test record to train a weight distribution model, predicts the optimal weight in the test phase, and generates a fusion feature vector by point multiplication of the dynamic weight and the feature vector.
[0007] In the model construction module, the four types of feature vectors after feature extraction are modeled using a graph neural network, a multi-modal fusion algorithm is used, and a cross-domain association graph and a global feature association matrix are obtained by quantizing the association strength. The specific operations are as follows: A cross-domain association graph is constructed between voltage, timing, and function, with four types of feature vectors, namely dynamic range, synchronization accuracy, performance indicators, and life prediction, as nodes. Edge weights are dynamically calculated through an attention mechanism; Four types of feature vectors are fused using an attention mechanism and a graph neural network. Four types of feature vector nodes are updated through a message passing mechanism to identify the transmission path of voltage fluctuations, timing deviations, and functional failures; The association strength between the four types of feature vectors is calculated through a graph convolution network, and a global feature association matrix is finally generated.
[0008] The intelligent scheduling unit comprises a splitting module and a fault prediction module, wherein: The splitting module: parses the splitting rule of the task module allocation scheme generated by the load balancing algorithm, and adjusts the dynamic parameters to obtain a subtask splitting table. The fault prediction module: combines voltage fluctuation and functional failure correlation data, uses voltage fluctuation, timing deviation and historical yield as three features to form a random tree model, inputs the voltage fluctuation value and timing deviation value into the random tree model, predicts the test failure risk of the current task, and dynamically adjusts the test strategy according to the predicted risk level.
[0009] The operation of the load balancing algorithm is as follows: The fusion feature vector is taken as the core input, and the performance index and dynamic range in the fusion feature vector are used to assign complexity weights to each task. According to the real-time running state of each functional module in the PXIe system, available modules are screened, and finally the available modules are combined with the weighted round robin and the least connection strategy to generate a task module allocation scheme.
[0010] The splitting rule of the task module allocation scheme generated by the load balancing algorithm in the splitting module is parsed, and the dynamic parameters are adjusted to obtain a subtask splitting table, which has the following specific operation: In the fusion feature vector, ADC test is divided into three levels of electrical, timing and function, which is regarded as a task splitting rule, and a subtask splitting scheme is generated by combining the historical templates of electrical specification verification, timing synchronization and functional verification. According to real-time test data, the subtask parameters are dynamically adjusted by an adaptive algorithm, and a subtask splitting table is formed by combining the subtask splitting scheme and the adjusted subtask parameters.
[0011] The adaptive test unit comprises a dynamic parameter feedback module and an adaptive module: The dynamic parameter feedback module: judges different threshold values, and then triggers different instructions, dynamically adjusts the parameters according to the corresponding trigger instructions, and obtains adjusted test parameters; The adaptive module: parses the weight of each dimension in the fusion feature vector, generates a test path according to the weight priority, verifies whether the generated test path covers high-risk fault points by combining voltage fluctuation and functional failure correlation data, automatically supplements test steps if there are vulnerabilities, and finally obtains adaptive test data.
[0012] In the dynamic parameter feedback module, different threshold values are judged, and different instructions are triggered, and the parameters are dynamically adjusted according to the corresponding trigger instructions to obtain adjusted test parameters, which has the following specific scheme: In the ADC test, the voltage fluctuation value is monitored in real time, and the detected voltage fluctuation is compared with the preset value: If the difference between the detected voltage and the preset value exceeds a, a voltage tolerance test step increase instruction is triggered; If the timing deviation is higher than b, that is, the timing deviation exceeds the tolerance, a sampling clock phase adjustment instruction is triggered; According to different trigger instructions, the test parameters are automatically adjusted: After receiving the voltage tolerance test step increase instruction, the sampling rate is automatically increased; After receiving the sampling clock phase adjustment instruction, the phase offset is automatically adjusted.
[0013] The intelligent safety test unit comprises a quantum encryption module and a high-precision measurement module: The quantum encryption module: a single-photon detector is used to establish a BB84 protocol QKD channel between the PXIe backplane and the edge FPGA module, and a quantum key is generated; A true random number seed is generated in combination with a quantum random number generator module, the test data is encrypted, and encrypted test data ciphertext is obtained; A quantum hash algorithm is integrated in the edge FPGA module, a hash value of the encrypted test data ciphertext is calculated in real time, and the PXIe high-speed bus is used to compare the hash value with a hash chain root node pre-stored in the cloud to obtain a millisecond-level integrity verification result: If the verification is successful, output: Hash matching is successful, and the data has not been tampered with; If the verification fails, output: Hash does not match, and trigger data exception alarm; The high-precision measurement module: a graphene nanometer sensor array is integrated in the PXIe test module, and then intelligent analysis is performed on the multi-physical field coupling to obtain high-precision measurement results and nanometer-level fault positioning information.
[0014] The high-precision measurement module integrates a graphene nanometer sensor array in the PXIe test module, and then performs intelligent analysis on the multi-physical field coupling to obtain high-precision measurement results and nanometer-level fault positioning information, and the specific method is as follows: Relying on cross-domain correlation model parameters, key features are screened through an attention mechanism, a coupling graph between electricity, heat and magnetism is constructed by using a graph neural network, and the influence of multi-physical field coupling effects on test accuracy is quantified, and picosecond-level timing accuracy, microvolt-level voltage resolution high-precision measurement results and nanometer-level fault points are output.
[0015] Compared with the prior art, the present application has the following advantages: The present application aims at the problem that test efficiency is limited by the number of stations, adopts a high-performance switch matrix to construct an instrument resource pool, realizes instrument resource allocation and routing driven by real-time demand of multiple stations in combination with a dynamic resource scheduler, cooperates with a parallel test architecture and a PXIe high-speed bus, significantly improves test throughput and reduces single-chip test cost, integrates a lightweight AI model to analyze chip intermediate results in real time during the test process, dynamically processes abnormal problems, realizes intelligent upgrading from passive error prevention to active diagnosis in combination with real-time monitoring of voltage fluctuation and timing deviation, adaptive parameter adjustment and a quantum encryption driven safety test module, and realizes real-time interaction of the test system and remote data through deep cooperation of a cloud platform, the local system is only responsible for high real-time test actions and data acquisition, the remote platform undertakes complex data processing, multi-parameter correlation modeling and historical trend prediction and visual report generation, combines a graph neural network cross-domain correlation model, a global feature correlation matrix and high-precision measurement fusion of a nanometer-level sensor array, forms a full-link closed loop from data acquisition to intelligent analysis, and finally realizes comprehensive upgrading of test efficiency, precision, flexibility, intelligence and data value, and meets the stringent requirements of semiconductor manufacturing on high-throughput, high-reliability diagnosis and full-link intelligent processing. BRIEF DESCRIPTION OF DRAWINGS
[0016] Fig. 1 A system diagram of the IC MultiSite automated test system based on PXIe of the present application; Fig. 2 A system architecture diagram of the IC MultiSite automated test system based on PXIe of the present application. DETAILED DESCRIPTION
[0017] The technical solutions in the embodiments of the present application will be clearly and completely described below in combination with the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application. EMBODIMENT
[0018] As shown in Figs. 1-2 The present embodiment provides an IC MultiSite automated test system based on PXIe, which includes a data fusion unit, an intelligent scheduling unit, an adaptive test unit and an intelligent safety test unit, wherein: The data fusion unit: acquires original data packets for feature extraction and mapping, constructs a cross-domain correlation model using four types of feature vectors after feature extraction, obtains three types of data, and obtains a fusion feature vector through dynamic weight distribution of the three types of data; The intelligent scheduling unit: through the load balancing algorithm, the fusion feature vector is adaptively task split, and the test record and data are fault predicted and avoided; The adaptive test unit: taking the original data packet for feature extraction and mapping, using the four types of feature vectors after feature extraction to construct cross The intelligent safety test unit: quantum encryption driving is performed on the adaptive test data output by the adaptive test unit, and a nanometer sensor array is introduced, and then high-precision measurement fusion is performed.
[0019] The data fusion unit includes a feature extraction module, a model construction module and a weight distribution module, wherein: The feature extraction module: acquires voltage range, current tolerance, clock jitter, ADC signal-to-noise ratio and high-temperature life four kinds of data as original data packet, respectively features electrical specifications, timing parameters, functional specifications and reliability data, to get dynamic range, synchronization accuracy, performance indicators and life prediction four types of feature vectors; Among them, for the calculation of dynamic range, S-type nonlinear normalization function is introduced to reflect the nonlinear influence of voltage fluctuation on system performance:
[0020]
[0021] The model construction module: uses the four types of feature vectors after feature extraction to establish a graph neural network correlation model, uses a multi-modal fusion algorithm on the correlation model, quantizes the correlation strength of the correlation model, and obtains a cross-domain correlation graph and a global feature correlation matrix; The weight distribution module: acquires real-time test data for real-time weight calculation, uses an adaptive weighting algorithm, combines a historical test record to train a weight distribution model, predicts an optimal weight in a test phase, and performs dot multiplication between the dynamic weight and the feature vector to generate a fusion feature vector.
[0022] It should be noted that the real-time test data includes real-time voltage fluctuation value, real-time current standard deviation value, real-time clock jitter value, real-time ADC signal-to-noise ratio value and real-time high-temperature life value; Among them, the calculation scheme of real-time weight calculation adopts a dynamic weight function:
[0023] In the model construction module, the four types of feature vectors after feature extraction are graph neural network correlation modeling, a multi-modal fusion algorithm is used, and a cross-domain correlation graph and a global feature correlation matrix are obtained by quantizing the correlation strength. The specific operation is as follows: A cross-domain correlation graph is constructed among voltage, timing and function, nodes are four types of feature vectors including dynamic range, synchronization accuracy, performance index and life prediction, and edge weights are dynamically calculated through an attention mechanism; Four types of feature vectors are fused through an attention mechanism and a graph neural network, four types of feature vector nodes are updated through a message passing mechanism, and a transmission path of voltage fluctuation, timing deviation and functional failure is identified; The correlation strength between the four types of feature vectors is calculated through a graph convolution network, and finally a global feature correlation matrix is generated.
[0024] The intelligent scheduling unit includes a splitting module and a fault prediction module, wherein: The splitting module: parses the splitting rule of the task module allocation scheme generated by the load balancing algorithm, and adjusts the dynamic parameters to obtain a sub-task splitting table; The fault prediction module: combines voltage fluctuation and functional failure correlation data, forms a random tree model using voltage fluctuation, timing deviation and historical yield as three features, inputs the voltage fluctuation value and timing deviation value into the random tree model, predicts the test failure risk of the current task, and dynamically adjusts the test strategy according to the predicted risk level.
[0025] The operation of the load balancing algorithm is as follows: The fusion feature vector is used as the core input, and the complexity weight of each task is allocated based on the performance index and dynamic range in the fusion feature vector; According to the real-time running state of each functional module in the PXIe system, the available modules are screened, and finally the available modules are combined with the weighted round robin and the least connection strategy to generate a task module allocation scheme.
[0026] The load balancing algorithm uses a dynamic load balancing score formula, which has the following specific form:
[0027] In the fusion feature vector, the ADC test is divided into three levels of electrical, timing and function, which is regarded as a task splitting rule, and a sub-task splitting scheme is generated by combining the historical templates of electrical specification verification, timing synchronization and functional verification; According to the real-time test data, the sub-task parameters are dynamically adjusted through an adaptive algorithm, and the sub-task splitting table is formed by combining the sub-task splitting scheme with the adjusted sub-task parameters.
[0028] The adaptive test unit includes a dynamic parameter feedback module and an adaptive module: The dynamic parameter feedback module: judges different threshold values, and then triggers different instructions, dynamically adjusts the parameters according to the corresponding trigger instructions, and obtains the adjusted test parameters; The adaptive module: analyze the weight of each dimension in the fusion feature vector, generate test path according to the weight priority, combine voltage fluctuation and functional failure correlation data, verify whether the generated test path covers high-risk fault points, if there is a loophole, automatically supplement the test steps, and finally get adaptive test data.
[0029] In this design, automatic supplement of test steps needs to supplement corresponding tests for specific fault points, which includes the following aspects: For targeted supplement of fault mode library, if the dynamic range feature in the fusion feature vector and the historical fault mode library show high risk of ADC linearity failure, automatically supplement the linearity scan range expansion and increase the non-linear error quantization test to ensure coverage of all range distortion points. When the timing deviation triggered by the synchronization accuracy feature is greater than 120ps, supplement the multi-clock domain phase calibration test, including dynamic adjustment of sampling clock phase and eye diagram analysis to verify the timing synchronization accuracy < 10ns.
[0030] For intelligent supplement of path optimization verification, verify the dynamically generated test path through the historical fault mode library, and judge whether it covers high-risk fault points in the order of high-temperature life test, ADC signal-to-noise ratio verification and linearity scan. If there is a loophole, automatically supplement fault injection test and fault recurrence rate verification to ensure that the path covers all high-risk fault modes.
[0031] In the dynamic parameter feedback module, different thresholds are judged to trigger different instructions, and the parameters are dynamically adjusted according to the corresponding trigger instructions to obtain the adjusted test parameters, and the specific scheme is as follows: In ADC test, real-time monitoring of voltage fluctuation value, and comparison of detected voltage fluctuation with preset value: If the difference between the detected voltage and the preset value exceeds a, trigger the voltage tolerance test step increase instruction; If the timing deviation is higher than b, that is, the timing deviation exceeds the tolerance, trigger the sampling clock phase adjustment instruction; Wherein, a is ±0.3V, b is 50ps; According to different trigger instructions, automatically adjust the test parameters: After receiving the voltage tolerance test step increase instruction, automatically increase the sampling rate; After receiving the sampling clock phase adjustment instruction, automatically adjust the phase offset.
[0032] It needs to be supplemented that when dynamically adjusting the parameters, the ADC voltage fluctuation value needs to be monitored in real time, and when it exceeds the threshold ±0.3V, the voltage tolerance test step is automatically supplemented, that is, the sampling rate is dynamically increased from 1GS / s to 1.2GS / s, and the power supply noise immunity test is increased to verify the signal integrity under ±1V fluctuation.
[0033] The intelligent safety test unit comprises a quantum encryption module and a high-precision measurement module: The quantum encryption module: a single-photon detector is used to establish a BB84 protocol QKD channel between a PXIe backboard and an edge FPGA module, and quantum keys are generated; A true random number seed is generated in combination with a quantum random number generator module, test data is encrypted, and encrypted test data ciphertext is obtained; A quantum hash algorithm is integrated in the edge FPGA module, and a hash value of the encrypted test data ciphertext is calculated in real time, the hash value is compared with a hash chain root node pre-stored in the cloud through a PXIe high-speed bus, and a millisecond-level integrity verification result is obtained: If the verification is successful, output: Hash matching is successful, and the data is not tampered with; If the verification fails, output: Hash does not match, and trigger data exception alarm; The quantum random number generator module is a core component for generating high-quality random numbers based on quantum physics principles, mainly used in high-security demand scenarios such as encrypted communication and key management, and is used for key management in the embodiment.
[0034] The high-precision measurement module: a graphene nanometer sensor array is integrated in a PXIe test module, and then intelligent analysis is performed on multi-physical field coupling to obtain high-precision measurement results and nanometer-level fault positioning information.
[0035] The high-precision measurement module integrates a graphene nanometer sensor array in a PXIe test module, and then performs intelligent analysis on multi-physical field coupling to obtain high-precision measurement results and nanometer-level fault positioning information, and the specific method is as follows: Relying on cross-domain association model parameters, key features are screened through an attention mechanism, a coupling graph between electricity, heat and magnetism is constructed using a graph neural network, and the influence of multi-physical field coupling effects on test accuracy is quantified, and picosecond-level timing accuracy, microvolt-level voltage resolution high-precision measurement results are output and nanometer-level fault points are located.
[0036] In the description of the present specification, the description of the terms "one embodiment", "example", "specific example" and the like means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are contained in at least one embodiment or example of the present application. In the present specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.
[0037] The preferred embodiments of the application disclosed above are only to facilitate the elucidation of the application. The preferred embodiments do not describe all the details of the application and limit the application to the specific embodiments described. Obviously, many modifications and variations can be made in light of the teachings above. The description is chosen and described in order to best explain the principles of the application and its practical application to thereby enable others skilled in the art to best utilize the application and get the best results from the application. The application is only limited by the claims and their full scope and equivalents.
Claims
1. An IC MultiSite automated test system based on PXIe, characterized in that, The intelligent safety testing unit comprises a data fusion unit, an intelligent scheduling unit, an adaptive testing unit and an intelligent safety testing unit, wherein: The data fusion unit: acquires original data packets for feature extraction and mapping, constructs a cross-domain correlation model using the four types of feature vectors after feature extraction, obtains three types of data, and obtains a fusion feature vector by dynamically allocating weights to the three types of data; The intelligent scheduling unit: performs adaptive task splitting on the fusion feature vector through a load balancing algorithm, and performs fault prediction and avoidance on the test records and data; The adaptive testing unit: dynamically adjusts the parameters of the test data and the fusion feature vector, and generates adaptive test data based on the fusion feature vector; The intelligent safety testing unit: performs quantum encryption driving on the adaptive test data output by the adaptive testing unit, and introduces a nanometer-level sensor array to obtain nanometer-level fault positioning information.
2. The PXIe-based IC MultiSite automated test system of claim 1, wherein, The data fusion unit comprises a feature extraction module, a model construction module and a weight allocation module, wherein: The feature extraction module: acquires four types of data, namely voltage range, current tolerance, clock jitter, ADC signal-to-noise ratio and high-temperature life, as original data packets, respectively features the electrical specifications, timing parameters, functional specifications and reliability data to obtain four types of feature vectors, namely dynamic range, synchronization accuracy, performance indicators and life prediction; The model construction module: uses the four types of feature vectors after feature extraction to establish a graph neural network correlation model, uses a multi-modal fusion algorithm on the correlation model, quantizes the correlation strength of the correlation model, and obtains a cross-domain correlation graph and a global feature correlation matrix; The weight allocation module: acquires real-time test data for real-time weight calculation, uses an adaptive weighting algorithm, combines a historical test record to train a weight allocation model, predicts the optimal weight in the test phase, and generates a fusion feature vector by multiplying the dynamic weight with the feature vector.
3. The PXIe-based IC MultiSite automated test system of claim 2, wherein, In the model construction module, the four types of feature vectors after feature extraction are graph neural network correlation modeled, a multi-modal fusion algorithm is used, and a cross-domain correlation graph and a global feature correlation matrix are obtained by quantizing the correlation strength, which is specifically as follows: A cross-domain correlation graph between voltage, timing and function is constructed, the nodes are four types of feature vectors, namely dynamic range, synchronization accuracy, performance indicators and life prediction, and the edge weights are dynamically calculated through an attention mechanism; Four types of feature vectors are fused through an attention mechanism and a graph neural network, four types of feature vector nodes are updated through a message passing mechanism, and the transmission path of voltage fluctuation, timing deviation and functional failure is identified; The correlation strength between the four types of feature vectors is calculated through a graph convolution network, and a global feature correlation matrix is finally generated.
4. The PXIe-based IC MultiSite automated test system of claim 1, wherein, The intelligent scheduling unit comprises a splitting module and a fault prediction module, wherein: The splitting module: analyzes the splitting rules of the task module distribution scheme generated by the load balancing algorithm, adjusts the dynamic parameters, and obtains a sub-task splitting table; The fault prediction module: in combination with voltage fluctuation and function failure correlation data, a random tree model is formed by using three characteristics of voltage fluctuation, timing deviation and historical yield, voltage fluctuation values and timing deviation values are input into the random tree model, test failure risks of the current task are predicted, and test strategies are dynamically adjusted according to the predicted risk levels.
5. The PXIe-based IC MultiSite automated test system of claim 4, wherein, The operation of the load balancing algorithm is as follows: The fusion feature vector is taken as a core input, and the performance index and dynamic range in the fusion feature vector are used to assign complexity weights to each task; According to the real-time running state of each functional module in the PXIe system, the available modules are screened, and finally the available modules are combined with the weighted round robin and the least connection strategy to generate a task module allocation scheme.
6. The PXIe-based IC MultiSite automated test system of claim 4, wherein, The splitting module analyzes the splitting rule of the task module allocation scheme generated by the load balancing algorithm, and adjusts the dynamic parameters to obtain a subtask splitting table, and the specific operation is as follows: In the fusion feature vector, ADC test is divided into three levels of electricity, timing and function, which is regarded as a task splitting rule, and a subtask splitting scheme is generated in combination with the historical templates of electrical specification verification, timing synchronization and function verification; According to real-time test data, the subtask parameters are dynamically adjusted through an adaptive algorithm, and the subtask splitting scheme is combined with the adjusted subtask parameters to form a subtask splitting table.
7. The PXIe-based IC MultiSite automated test system of claim 1, wherein, The adaptive test unit includes a dynamic parameter feedback module and an adaptive module: The dynamic parameter feedback module: different thresholds are judged, different instructions are triggered, parameters are dynamically adjusted in combination with the corresponding trigger instructions, and adjusted test parameters are obtained; The adaptive module: analyze the weight of each dimension in the fusion feature vector, generate a test path according to the weight priority, verify whether the generated test path covers high-risk fault points in combination with voltage fluctuation and function failure correlation data, automatically supplement test steps if there are vulnerabilities, and finally obtain adaptive test data.
8. The PXIe-based IC MultiSite automated test system of claim 7, wherein, In the dynamic parameter feedback module, different thresholds are judged, different instructions are triggered, parameters are dynamically adjusted in combination with the corresponding trigger instructions, and adjusted test parameters are obtained, and the specific scheme is as follows: In ADC test, the voltage fluctuation value is monitored in real time, and the detected voltage fluctuation is compared with the preset value: If the difference between the detected voltage and the preset value exceeds a, the voltage tolerance test step increase instruction is triggered; If the timing deviation is higher than b, that is, the timing deviation exceeds the tolerance, the sampling clock phase adjustment instruction is triggered; According to different trigger instructions, the test parameters are automatically adjusted: After receiving the voltage tolerance test step increase instruction, the sampling rate is automatically increased; After receiving the sampling clock phase adjustment instruction, the phase offset is automatically adjusted.
9. The PXIe-based IC MultiSite automated test system of claim 1, wherein, The intelligent safety test unit includes a quantum encryption module and a high-precision measurement module: The quantum encryption module: a BB84 protocol QKD channel is established between the PXIe backplane and the edge FPGA module through a single-photon detector to generate a quantum key; In combination with the quantum random number generator module, a true random number seed is generated, test data is encrypted, and encrypted test data ciphertext is obtained; Integrate quantum hash algorithm in edge FPGA module, calculate hash value of encrypted test data ciphertext in real time, compare hash value with pre-stored hash chain root node in cloud through PXIe high-speed bus, and obtain millisecond-level integrity verification result: If the verification is successful, output: Hash matching success, data has not been tampered with; If the verification fails, output: Hash mismatch, trigger data exception alarm; The high-precision measurement module: integrate graphene nanometer sensor array in the PXIe test module, then intelligently analyze the multi-physical field coupling to obtain high-precision measurement results and nanometer-level fault positioning information.
10. The PXIe-based IC MultiSite automated test system of claim 9, wherein, The high-precision measurement module integrates graphene nanometer sensor array in the PXIe test module, then intelligently analyzes the multi-physical field coupling to obtain high-precision measurement results and nanometer-level fault positioning information, and the specific method is as follows: Relying on cross-domain correlation model parameters, filter key features through attention mechanism, use graph neural network to construct coupling atlas among electricity, heat and magnetism, and quantify the influence of multi-physical field coupling effect on test accuracy, output high-precision measurement results of picosecond-level timing accuracy and microvolt-level voltage resolution and locate nanometer-level fault points.