Operating Mode Detection in Multi-Modal Assets With Noisy Time-Series
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing Gaussian mixture models (GMM) and Gaussian Graphical models (GGM) struggle to handle multi-modality and multi-assets, leading to inaccurate identification of operating modes due to noisy data and disregard of individual asset characteristics.
Innovation Solution
A Mode Fidelity Mixture (MFM) model is developed to learn a customized prediction function for each individual asset, leveraging a graphical mixture model to capture both commonality and individuality of assets, while handling multi-modality and noisy data through sparse Gaussian graphical models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional Gaussian mixture models are used to handle multi-modality, then operating modes can be identified, but individual asset characteristics are disregarded leading to inaccurate identification
Solution Approach 1:
The patent segments the single GMM model into multiple asset-specific GMM models, where each asset has its own customized prediction function. This segmentation allows each model to capture individual asset characteristics while maintaining the ability to handle multi-modality, thereby resolving the contradiction between adaptability and measurement precision.
Solution Approach 2:
The patent applies local quality by allowing each asset to have its own customized prediction function with asset-specific parameters. Instead of using a uniform model for all assets, the system tailors the model characteristics to match local asset properties, improving operating mode identification accuracy while preserving multi-modality handling capabilities.
2Measurement precision
If traditional Gaussian Graphical models are used for single-modal assets, then individual asset characteristics are captured, but multi-modality cannot be handled
Solution Approach 1:
The patent merges the strengths of GGMs (capturing asset individuality through sparse precision matrices) with the multi-modality handling capability of GMMs. By combining these approaches into a unified framework where each asset has its own GGM-based GMM model, the system simultaneously achieves both asset-specific modeling and multi-mode detection.
3Measurement precision
If customized prediction functions are learned for each individual asset, then operating mode identification accuracy improves, but computational complexity increases
Solution Approach 1:
The patent extracts and enforces the sparsity structure from GGMs into the asset-specific GMM models. By taking out only the essential sparse dependency relationships and applying them to each asset's model, the system reduces the number of parameters that need to be learned, thereby lowering computational complexity while maintaining high identification accuracy.
Solution Approach 2:
The patent changes the parameter structure by imposing sparsity constraints on the precision matrices of each asset's GMM model. This parameter change reduces the effective number of parameters from O(d^2) to O(d), where d is the dimensionality, significantly reducing computational complexity while preserving the ability to capture asset individuality.
Data Source
AI summary
A system and method for learning a predictive function that can automatically learn different operating modes for a multi-modal system and predict the number of operating states for a multi-modal system and additionally the detailed structure for each state. Once learned, the predictive function (model) can be used to determine a mode of a new sample (an asset). Based on the determined components that maximize a log likelihood function, a mode of the new sample is detected into the model via dependency graphs. One aspect includes enforcing a lower bound for the number of sample points to form an operational mode for an asset. While a mode relates to sample points which maximizes like log-likelihood, an ability is provided to remove artifact modes due to noisy data by considering a sufficient sample data condition and maximizing log-likelihood. Domain knowledge can be incorporated into the model via dependency graphs.