Machine Learning Model Predicts Failing Tests for Devices Under Test
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Automatic test equipment (ATE) faces challenges in predicting which tests a device under test (DUT) will fail, as existing methods lack efficiency and accuracy in identifying failing tests, especially in high-speed and complex testing scenarios.
Innovation Solution
A machine learning model is trained using data from initial sets of DUTs to predict which tests will fail in a different set of DUTs with similar features, allowing for pattern matching and real-time prediction during testing, enabling the ATE to continue testing and analyze results for failure causes without test insertion, with the model capable of executing in under 10 milliseconds.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional testing methods are used to identify failing tests, then testing coverage is maintained, but prediction accuracy and efficiency are insufficient
Solution Approach 1:
The system performs preliminary training using historical test data from an initial set of DUTs to build a machine learning model before actual testing. This pre-computed model enables rapid prediction during production testing, eliminating the need for real-time analysis of all test results and significantly improving testing efficiency while maintaining high prediction accuracy.
Solution Approach 2:
The system creates a computational model that replicates the testing behavior of the initial DUT set. By copying the failure patterns and test result structures from the training set into a machine learning model, the system can predict outcomes for new DUTs without physically re-testing all parameters, thereby improving both accuracy and efficiency.
2Reliability
If comprehensive testing is performed on all DUTs, then diagnostic capability is maintained, but testing time and latency increase
Solution Approach 1:
The system extracts only the critical failure patterns and test result features from comprehensive testing of the initial DUT set. By taking out only the essential diagnostic information needed for prediction rather than processing all test data in real-time, the system maintains diagnostic capability while reducing testing latency to under 10 milliseconds.
Solution Approach 2:
The system performs preliminary extraction and analysis of diagnostic features during the training phase. This pre-processing of diagnostic information enables the model to make rapid predictions during production testing without sacrificing diagnostic capability, as the critical diagnostic patterns have already been identified and encoded in the model.
3Productivity
If machine learning prediction is implemented in real-time, then testing speed improves, but model training complexity increases
Solution Approach 1:
The system segments the machine learning implementation into two distinct phases: an offline training phase where the model is built using historical data, and an online prediction phase where the trained model is applied in real-time testing. This segmentation allows complex model training to occur separately from production testing, enabling fast real-time predictions without compromising on model training capabilities.
Data Source
AI summary
Example techniques may be implemented as a method, a system or more non-transitory machine-readable media storing instructions that are executable by one or more processing devices, Operations performed by the example techniques include obtaining data representing results of tests executed by one or more test instruments on an initial set of devices under test (DUTs) in a test system; and using the data to train a machine learning model. The machine learning model is for predicting which of the tests will produce failing results for a different set of DUTs. DUTs in the different set have one or more features in common with DUTs in the initial set.