Phase Space Fault Diagnostics Across Variable Operating Conditions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional fault diagnostic techniques face challenges in accurately diagnosing the nature and magnitude of faults in dynamic systems due to their non-linear nature, requiring complex machine learning models and offline system disassembly for inspection, and are often limited to specific operating conditions.
Innovation Solution
A data processing apparatus and method that acquires signals from dynamic systems, determines phase spaces by correlating signals, and uses density-based, embedded dimension, or digital signal phase space topology processes to extract features and train fault detection models, enabling diagnostic testing across varying operating conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional fault diagnostic techniques are used, then fault detection is possible, but the system requires offline disassembly and is limited to specific operating conditions
Solution Approach 1:
The patent replaces mechanical disassembly with signal processing-based diagnostics. Phase space reconstruction and topology analysis allow fault detection through operational signals alone, eliminating the need for physical system disassembly while maintaining diagnostic capability across varying operating conditions
Solution Approach 2:
The patent transforms operating condition parameters into diagnostic advantages by using phase space reconstruction that adapts to different operating states. The method extracts topological features that remain discriminative across speed and load variations, converting parameter variability from a limitation into a diagnostic resource
2Measurement precision
If conventional machine learning models are used, then fault detection capability is improved, but model complexity and processing power requirements increase
Solution Approach 1:
The patent extracts essential topological features from phase space representations that directly characterize fault states. By focusing on topological invariants and geometric properties rather than full signal complexity, the method achieves high detection accuracy with simplified feature sets, reducing model complexity while maintaining precision
Solution Approach 2:
The patent transitions from time-domain signal analysis to phase space dimensionality, where faults manifest as distinct topological structures. This dimensional transformation reveals inherent geometric patterns that simplify classification, enabling accurate fault detection with less complex models by operating in a more informative feature space
3Reliability
If conventional pattern recognition techniques are used, then fault detection works for defined operating conditions, but the model must be retrained when operating conditions change
Solution Approach 1:
The patent creates a universal diagnostic framework based on phase space topology that functions across multiple operating conditions without retraining. The topological features extracted from reconstructed phase spaces are invariant to operational variations, allowing a single model to reliably detect faults across different speeds, loads, and system configurations
Solution Approach 2:
The patent employs dynamic phase space reconstruction that adapts to changing operating conditions in real-time. By continuously reconstructing phase space from incoming signals and extracting topological features, the system maintains diagnostic reliability across varying conditions without requiring model retraining, as the topological structure naturally adapts to the current operational state
Data Source
AI summary
A computer-implemented method is provided that includes acquiring signals characterizing an operation condition of a first component of a first dynamic system, determining a phase space of the dynamic system by correlating the acquired signals and determining at least one of a time lag and mutual information of the correlated signals, determining features of the phase space of the acquired signals based on performing one of a density based process, an embedded dimension process, or a digital signal phase space topology process, training a fault detection model based on the determined features, and performing, based on the trained fault detection model, diagnostic testing on a second component of a second dynamic system to detect an operation condition fault of the second component.