ML Anomaly Detection in DUT Test Measurement Results
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing test and measurement processes for semiconductor chips are time-consuming and resource-intensive due to the manual analysis of large datasets, lacking effective automation and real-time feedback, which increases costs and delays in identifying anomalies.
Innovation Solution
A machine learning and artificial intelligence-based system that analyzes consolidated test and measurement results in real-time, using anomaly detection models to identify outliers and provide near-real-time feedback, reducing the need for manual analysis and accelerating the detection of anomalies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional manual analysis methods are used to analyze test and measurement data, then measurement precision can be maintained, but loss of time increases significantly
Solution Approach 1:
The patent replaces manual mechanical analysis with an automated machine learning system. The ML model automatically processes test and measurement data, identifying anomalies without human intervention. This substitution maintains detection accuracy while dramatically reducing the time required for post-analysis, as the system can process large datasets instantly compared to manual review processes.
Solution Approach 2:
The system enables self-service anomaly detection where the test equipment automatically performs data analysis and anomaly identification without requiring expert human analysis. The ML model is trained to independently detect anomalies, allowing the system to serve itself in the analysis process, thereby eliminating time-consuming manual review while maintaining precision.
2Reliability
If comprehensive test validation is performed on semiconductor chips, then reliability improves, but loss of time and manufacturing costs increase
Solution Approach 1:
The patent implements preliminary anomaly detection during the test execution phase itself, rather than waiting for post-test analysis. The ML model continuously monitors test data in real-time and flags potential anomalies as they occur. This preliminary action allows validation to be more efficient, maintaining high reliability while reducing overall validation cycle time by identifying issues earlier in the process.
Solution Approach 2:
The system provides real-time feedback during test execution, allowing immediate identification and correction of anomalies. The ML model continuously analyzes incoming test data and provides feedback about potential issues, enabling dynamic adjustment of test parameters or immediate failure identification. This feedback mechanism maintains high validation quality while reducing time losses associated with delayed anomaly detection.
3Measurement precision
If extensive manual analysis of test data is performed, then measurement precision is maintained, but device complexity increases due to resource requirements
Solution Approach 1:
The patent replaces complex manual analysis processes with a streamlined machine learning system. Rather than requiring multiple human analysts with specialized tools and procedures, a single trained ML model performs the same function with simpler infrastructure requirements. The model maintains measurement precision while reducing the operational complexity of the analysis system.
4Productivity
If real-time anomaly detection is implemented using ML/AI, then productivity increases, but device complexity increases due to implementation requirements
Solution Approach 1:
The patent implements a universal ML platform that can handle multiple test types, measurement protocols, and anomaly detection scenarios through a single system architecture. The ML model is designed to be multi-functional, adapting to different semiconductor test requirements without requiring separate specialized systems. This universality increases productivity across various test scenarios while managing device complexity through code reusability and standardized interfaces.
Data Source
AI summary
A test and measurement device has an interface, one or more connectors, each connector to allow the test and measurement device to connect to a test and measurement instrument, and one or more processors, the one or more processors configured to execute code to cause the one or more processors to: receive one or more user inputs through the interface identifying one or more tests to perform on a device under test (DUT); form a connection through one of the one or more connectors to the DUT to perform the one or more tests and receive test result data; apply one or more machine learning models to the test result data to identify potentially anomalous test results; and generate and present a representation of the test result data and the potentially anomalous test results. A method of analyzing test data includes receiving one or more user inputs through an interface identifying one or more test to perform on a device under test (DUT), forming a connection to at least one test and measurement instrument, directing the test and measurement instrument to perform one or more tests on the DUT and receive test result data, applying one or more machine learning models to the test result data to identify potentially anomalous test results, and generating and presenting a representation of the test result data and the potentially anomalous test results.