Virtual Sensor Fault Classification With Fewer Physical Sensors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing apparatuses face challenges in efficiently detecting and isolating fault conditions using a limited number of sensors, which can lead to increased costs and weight, while employing a large number of sensors results in insufficient insight into the cause of the fault.
Innovation Solution
A method involving a control circuit that processes sensor information to output virtual parameters, allowing for the identification of fault classes rather than individual failure modes, using a reduced number of sensors and potentially incorporating a neural network trained with data from similar or identical apparatuses.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a large number of sensors are employed to monitor apparatus failures, then fault detection capability is improved, but device complexity and cost increase
Solution Approach 1:
The patent introduces virtual parameters as an intermediary layer between physical sensors and fault diagnosis. These virtual parameters are calculated from sensor data and represent complex system states without requiring direct physical measurement, thereby reducing sensor count while maintaining diagnostic capability
Solution Approach 2:
The patent replaces physical sensors with virtual sensing through computational models. Neural networks and diagnostic algorithms process existing sensor data to generate virtual sensor readings, substituting mechanical sensing elements with information processing mechanisms
2Device complexity
If a limited number of sensors are used to reduce complexity, then device complexity is reduced, but measurement precision and fault insight deteriorate
Solution Approach 1:
The patent transforms physical sensor parameters into virtual parameters through computational relationships. By changing the parameter representation from direct physical measurements to derived virtual quantities, the system achieves enhanced measurement precision with fewer sensors
Solution Approach 2:
The patent creates composite information structures by combining data from multiple sensors through neural networks and diagnostic algorithms. This composite processing extracts deeper fault insights from limited sensor inputs, analogous to how composite materials achieve enhanced properties
3Device complexity
If virtual parameters are used to represent apparatus states, then device complexity is reduced, but information processing requirements increase
Solution Approach 1:
The patent performs preliminary computational processing by pre-training neural networks with historical data and pre-calculating virtual parameter relationships. This preliminary action reduces real-time computational energy requirements during actual monitoring operations
Data Source
AI summary
Sensor information from a plurality of sensors that monitor an apparatus is accessed and then sensed information that corresponds to that sensor information is input into a control circuit configured to output virtual parameters corresponding to the apparatus as a function, at least in part, of the sensed information. A particular fault class is determined from amongst a plurality of fault classes as a function, at least in part, of the virtual parameters and at least one task is identified regarding the apparatus that corresponds to the particular fault class.