Automatic sensor trace validation using machine learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing sensor validation systems in immunoassay and clinical chemistry analyzers face challenges in reliably detecting device failures due to ambient noise, leading to complex and difficult-to-maintain rule engines, which can result in frequent false alarms and maintenance needs.
Innovation Solution
A computer-implemented method using a machine learning model, specifically a deep neural network, to encode and analyze sensor trace signals, reconstruct signals, and estimate parameters, thereby identifying valid or invalid signals and categorizing failures, while being invariant to noise and highlighting discrepancies in a graphical user interface.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If rule engines are used to validate sensor traces with safety margins, then device failure detection is achieved, but the system becomes increasingly complex and difficult to maintain due to ambient noises
Solution Approach 1:
The patent replaces the mechanical rule-based validation system with a machine learning model that automatically learns validation patterns from training data. The ML model substitutes complex if-else rule engines with a trained neural network that processes sensor traces and automatically identifies failures, eliminating the need for manually crafted safety margin rules while maintaining detection accuracy.
Solution Approach 2:
The patent transforms the validation approach by changing from fixed safety margin parameters to dynamic thresholds learned from training data. The ML model adapts its decision boundaries based on patterns in the training set, allowing it to handle ambient noises without requiring explicit rule adjustments for each noise scenario.
2Reliability
If rule engines handle variations caused by ambient noises, then validation robustness improves, but maintenance difficulty increases
Solution Approach 1:
The patent applies preliminary action by training the machine learning model in advance on a comprehensive dataset that includes various ambient noise conditions. This pre-training enables the model to learn robust validation patterns that generalize to unseen noise scenarios, eliminating the need for ongoing rule adjustments and maintenance when new noise patterns emerge.
Solution Approach 2:
The ML model performs self-service by automatically adapting to ambient noises through its trained patterns without requiring manual rule updates. The system maintains itself by relying on the generalization capabilities of the trained model rather than requiring operators to continuously adjust validation rules for new noise conditions.
3Reliability
If safety margins are used for automated validation, then false alarms are reduced, but the system becomes less adaptable to new failure patterns
Solution Approach 1:
The patent introduces dynamics by replacing static safety margin thresholds with a dynamic machine learning model that can adapt to new failure patterns. The ML model's decision boundaries are not fixed but are determined by the training data, allowing the system to naturally adapt to new failure modes as they appear in the training set, while still maintaining low false alarm rates through learned patterns.
Data Source
AI summary
A computer-implemented method for detecting a failure of a device connected to a sensor is disclosed. The method includes a machine learning model receiving a trace signal from the sensor indicating a status of the device, the machine learning model encoding the trace signal into a plurality of vector representations, and the machine learning model determining whether the trace signal is valid or invalid based on the plurality of vector representations.