Virtual Sensing for Predictive Reliability in Hostile Environments
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing predictive maintenance systems for machinery and aerospace vehicles rely on real-time sensor data, which can be of poor quality and lead to reactionary maintenance practices, failing to adequately capture the need for maintenance due to delayed detection and faulty readings.
Innovation Solution
Implementing virtual sensing using artificial intelligence and deep neural networks to generate virtual sensors that estimate measurements and operating conditions, particularly in hostile environments where physical sensors are expensive or impractical, by training the networks with sensor data and plume image data to determine shock wave data and parameters like Mach number and internal pressure.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If real-time physical sensors are used for predictive maintenance, then maintenance can be performed, but the sensor data quality is poor leading to delayed detection and reactionary maintenance
Solution Approach 1:
The patent creates virtual copies of physical sensors through deep neural networks that process available sensor data to generate synthetic measurements. These virtual sensors replicate the function of physical sensors without requiring direct physical contact with the monitored system, thereby improving measurement quality while maintaining predictive maintenance capability
Solution Approach 2:
The patent replaces physical mechanical sensors with an information-processing-based virtual sensing system. Instead of using physical sensors that directly contact the system and suffer from quality issues, the system uses deep neural networks to process existing sensor data and generate virtual sensor readings, substituting mechanical sensing with computational sensing
2Measurement precision
If physical sensors are deployed in hostile environments, then measurements can be obtained, but the cost is prohibitively expensive or performance degrades
Solution Approach 1:
The patent creates virtual sensor copies that can operate in hostile environments without physical deployment. By using deep neural networks to generate virtual sensor data from available measurements, the system obtains measurement capability in environments where physical sensors would be too expensive or degrade performance, eliminating the need for costly physical sensor deployment in harsh conditions
3Reliability
If conservative maintenance practices are used due to poor sensor quality, then failures are avoided, but downtime increases and maintenance costs rise
Solution Approach 1:
The patent implements a feedback mechanism where virtual sensor data continuously monitors system health and feeds into predictive maintenance algorithms. This closed-loop system provides accurate real-time feedback on system condition, enabling maintenance to be performed only when actually needed rather than following conservative schedules, thereby reducing unnecessary downtime while maintaining reliability
Solution Approach 2:
The patent enables preliminary detection of potential failures through virtual sensing that identifies degradation trends before actual failures occur. By detecting early signs of problems through processed sensor data and virtual sensor readings, the system allows maintenance to be planned and executed at optimal times rather than reacting to failures or following overly conservative schedules
Data Source
AI summary
An exemplary virtual sensing method and system are disclosed for predictive reliability (VIPR) procedure and/or controls that employ artificial intelligence and machine learning (AI/ML), particularly deep neural networks and multi-modal deep learning, with vehicle sensor data to create virtual sensors. The virtual sensors can be used to estimate measurements and operating conditions in a hostile environment in rockets and vehicle systems.