Satellite Predictive Maintenance via Dual-Model Lifespan Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current satellite management systems lack effective methods for predicting and preventing failure events, leading to potential catastrophic failures such as complete shutdown, fire, or total communication loss, due to the complexity of disambiguating failure modes from noisy sensor data streams.
Innovation Solution
A method involving a computer system that accesses historical datasets, segments failure events, extracts features, and trains a lifespan prediction model using adversarial or convolutional neural networks to predict satellite lifespan, allowing for early detection of impending failures and prompting operators for corrective actions through an operator portal.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional satellite monitoring systems are used, then operational simplicity is maintained, but failure prediction capability is insufficient leading to catastrophic failures
Solution Approach 1:
The patent segments failure prediction into two distinct models: a long-term lifespan prediction model and a short-term failure probability model. This segmentation allows each model to specialize in different time horizons, improving overall prediction accuracy without requiring a single complex system to handle all scenarios.
Solution Approach 2:
The patent introduces a temporal dimension by analyzing satellite data across different time windows (long-term vs. short-term). This dimensional approach transforms the failure prediction problem from a static assessment to a dynamic, time-aware system that can distinguish between gradual degradation and imminent failure.
2Measurement precision
If detailed failure mode analysis is attempted, then diagnostic precision is improved, but system complexity and difficulty of detection increase due to noisy sensor data
Solution Approach 1:
The patent extracts only the essential information needed for failure prediction by focusing on lifespan and failure probability metrics rather than attempting to fully classify and diagnose every possible failure mode. This extraction approach reduces the complexity of handling noisy sensor data while maintaining effective prediction capability.
Solution Approach 2:
Instead of trying to identify specific failure modes from noisy data, the patent inverts the approach by directly predicting failure probability and lifespan from the aggregate sensor data. This inversion bypasses the difficult intermediate step of failure mode classification while still achieving reliable failure prediction.
3Speed
If continuous monitoring at high frequency is implemented, then failure detection speed is improved, but energy consumption increases
Solution Approach 1:
The patent implements dynamic monitoring by adjusting the analysis frequency based on the predicted failure probability. When the short-term failure probability is high, the system increases monitoring intensity and alerts operators for immediate attention. When probability is low, monitoring can be reduced, optimizing energy usage while maintaining detection speed when needed.
Solution Approach 2:
The patent changes the parameter of monitoring intensity based on the predicted failure probability. Instead of continuous high-frequency monitoring, the system dynamically adjusts the level of monitoring and data analysis based on risk assessment, reducing energy consumption during low-risk periods while maintaining rapid detection capability when risks are elevated.
Data Source
AI summary
One variation of a method includes: training a first model to predict failures within the first population of satellites within a first time window based on a first set of historical timeseries telemetry data and a first set of historical timeseries failure data; and training a second model to predict failures within the first population of satellites within a second time window, shorter than the first time window, based on the first set of historical timeseries telemetry data and the first set of historical timeseries failure data. The method further includes: predicting a first probability of failure of the first satellite within the first time window based on the first model and the first set of timeseries telemetry data; and predicting a second probability of failure of the first satellite within the second time window based on the second model and the first set of timeseries telemetry data.