Faulty Device Detection via Near-Field to Far-Field Extrapolation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting faulty devices with wireless communication capabilities, such as those employing machine vision and machine learning, are inefficient and inaccurate due to purely optical investigations.
Innovation Solution
A method and system utilizing machine learning techniques to gather near-field test data, extrapolate it to far-field conditions, and evaluate device performance, reducing the number of measurement antennas and incorporating sparse test data, spherical wave expansion, and recalibration capabilities to detect faulty devices efficiently and accurately.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If machine vision and machine learning are used for product inspection, then automation is improved, but measurement precision deteriorates for wireless communication devices
Solution Approach 1:
The patent replaces optical inspection methods with electromagnetic field-based measurement. Instead of using cameras and visual algorithms to detect defects, the system uses antenna arrays to measure electromagnetic radiation patterns, which directly characterize the wireless communication performance of devices. This substitution enables both automation and high measurement precision for wireless devices.
Solution Approach 2:
The patent changes the measurement parameter from optical reflectivity/color (traditional machine vision) to electromagnetic radiation characteristics (far-field patterns). By measuring parameters such as radiation intensity, directionality, and polarization in the far field, the system achieves precise detection of wireless communication device performance and defects.
2Measurement precision
If far field measurement is performed to accurately evaluate device performance, then measurement precision is improved, but measurement time and resources increase
Solution Approach 1:
The patent performs preliminary near-field measurements during the manufacturing process and uses these measurements to predict far-field performance through machine learning models. This preliminary action allows quality assessment to be conducted early in production, eliminating the need for time-consuming separate far-field measurement campaigns while maintaining measurement precision.
Solution Approach 2:
The patent creates a virtual copy of the far-field measurement scenario through machine learning models trained on near-field data. Instead of physically performing expensive and time-consuming far-field measurements with large antenna arrays, the system uses computational models that replicate far-field radiation patterns from near-field measurements, achieving the same measurement precision with minimal time and resource investment.
3Measurement precision
If traditional measurement methods are used, then measurement accuracy is maintained, but device complexity and measurement setup requirements increase
Solution Approach 1:
The patent segments the measurement process into two distinct parts: simple near-field measurements using minimal antennas during manufacturing, and complex far-field simulation using machine learning models. This segmentation allows the actual physical measurement setup to be simple and low-complexity, while the computational complexity is handled separately through software algorithms, reducing overall measurement setup requirements.
Solution Approach 2:
The patent introduces machine learning models as an intermediary between near-field measurements and far-field performance evaluation. This intermediary component translates simple near-field antenna measurements into accurate far-field predictions, eliminating the need for complex far-field measurement hardware while maintaining measurement accuracy. The ML model acts as a computational bridge that simplifies the overall measurement system.
Data Source
AI summary
A method and a system for detecting faulty devices are provided. The method comprises the steps of gathering test data in near field with respect to a device under test, extrapolating the test data to far field conditions with the aid of at least one machine learning technique, and evaluating a far field performance of the device under test on the basis of the far field conditions.