Virtual Sensor Fault Detection Using Model-Based Sensor Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Virtual sensors in electronic devices are prone to errors due to intermittent or permanent malfunctions in hardware sensors, leading to erroneous behavior and potential malfunctions.
Innovation Solution
A software-based solution using machine learning and neural networks to detect and compensate for faulty sensor inputs by comparing sensor measurements against a predetermined model, adjusting the virtual sensor's operation to ignore or switch to alternative neural networks when necessary.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If virtual sensors use sensor fusion techniques to combine information from multiple physical sensors, then the device can create new sensor outputs and reduce dependency on specialized hardware sensors, but the system becomes vulnerable to erroneous behavior when at least one input sensor is malfunctioning
Solution Approach 1:
The system continuously monitors sensor outputs and compares them against expected patterns. When a sensor is detected to be malfunctioning, the feedback mechanism triggers switching to alternative sensors or virtual sensor modes, ensuring continuous reliable operation despite hardware failures
Solution Approach 2:
A virtual sensor layer acts as an intermediary between physical sensors and the final output. This virtual layer processes and validates sensor data, filtering out erroneous information from malfunctioning sensors while synthesizing reliable outputs from functioning sensors
2Reliability
If the system implements comprehensive sensor monitoring and fault detection mechanisms, then the robustness of virtual sensors is enhanced, but the device complexity and processing requirements increase
Solution Approach 1:
The system performs self-diagnosis by automatically monitoring its own sensor outputs for patterns indicating malfunction. The fault detection mechanisms are integrated into the existing sensor fusion pipeline, allowing the system to detect and respond to its own failures without external intervention
Solution Approach 2:
Instead of implementing comprehensive monitoring of all possible sensor failure modes, the system focuses on detecting the most common and critical sensor malfunctions. This selective approach provides adequate reliability protection while minimizing the added complexity of monitoring mechanisms
3Duration of action of stationary object
If the system switches between different neural networks or sensor combinations based on detected failures, then the continuity of device operation is maintained, but the processing time and energy consumption increase
Solution Approach 1:
The system pre-configures multiple neural networks and sensor combinations that can be activated in response to specific sensor failures. When a malfunction is detected, the system can quickly switch to a pre-prepared alternative without needing to retrain or reconstruct the processing pipeline from scratch
Solution Approach 2:
The system periodically monitors sensor health and proactively switches to alternative configurations before complete failures occur, preventing operational interruptions while minimizing the frequency of costly processing switches by detecting early signs of sensor degradation
Data Source
AI summary
The present invention relates to an electronic device and corresponding method, where the electronic device includes at least two sensors. The at least two sensors are in communication with a virtual sensor including a processing unit adapted to monitor the use of the device based on input from said sensors. The at least two sensors include activity sensors sensing user activities related to the device, and the processing unit is configured to receive and analyze the input from said sensors during use and to provide and store a model representing said typical use of the device. The device further being configured to compare new measurements received from said sensors with said model, providing a signal if the new measurements from said sensors deviates by a predetermined limit from the measurements predicted by the model.