Neural Graph Sensor Calibration for Ageing Compensation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Sensor systems face challenges in maintaining optimal calibration over time due to ageing and ambient influences, requiring recalibration but often lack sufficient data for effective recalibration, especially in end applications where test bench data is unavailable.
Innovation Solution
A method using a data-based calibration model trained as a neural graph network that adapts through unsupervised learning, creating a sensor state graph from detection and state variables to generate correction variables, allowing recalibration without labeled data by evaluating augmented graphs and calculating loss for retraining.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If recalibration is performed using traditional methods, then initial calibration accuracy is maintained, but the calibration model cannot adapt to ageing effects and ambient condition changes over time
Solution Approach 1:
The calibration model transitions from a static initial calibration to a dynamic adaptive system that continuously updates its parameters during operation. The neural graph network learns from operational data and recalibrates itself in response to ageing effects and environmental changes, making the calibration process dynamic rather than fixed.
Solution Approach 2:
The sensor system performs self-calibration using its own operational data and detected sensor states. The unsupervised learning mechanism enables the system to automatically adapt to changes without external intervention or labeled training data, effectively serving its own calibration needs during operation.
2Adaptability or versatility
If unsupervised retraining is performed without labeled data, then recalibration during operation is enabled, but the precision of calibration model adaptation is reduced
Solution Approach 1:
The sensor state graph serves as an intermediary representation that captures the complex relationships between sensor readings, environmental conditions, and sensor states. This structured graph representation enables the unsupervised learning process to effectively adapt the calibration model while maintaining precision by providing a meaningful framework for analyzing operational data.
3Measurement precision
If test bench calibration is used, then initial calibration precision is high, but recalibration cannot be performed during actual operation in end applications
Solution Approach 1:
The patent replaces the mechanical test bench calibration system with a software-based unsupervised learning mechanism. Instead of requiring physical test equipment and controlled calibration environments, the system uses computational algorithms that process operational sensor data to perform recalibration during actual use, substituting mechanical calibration infrastructure with intelligent software processing.
Data Source
AI summary
A computer-implemented method for recalibrating a trained data-based calibration model for use in a sensor system for measuring one or more physical variables is disclosed. The data-based calibration model is formed as a neural graph network which is trained to output an output vector comprising one or more output variables and a plurality of auxiliary variables, depending on a sensor state graph representing a sensor state. The method includes (i) detecting one or more detection variables relating to the one or more physical variables and one or more state variables indicating one or more environmental influences on the sensor system, (ii) ascertaining a sensor state graph depending on the one or more detection variables and the one or more state variables at a time of detection, (iii) augmenting the sensor state graph, (iv) evaluating the augmented sensor state graph with the data-based calibration model to obtain the output vector, (v) determining a loss depending on the output vector, and (vi) unsupervised training of the calibration model depending on the determined loss.