Inferential Model for Stair-Stepped Sensor Signal Interpolation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing prognostic-surveillance systems face poor cross-correlation and performance issues due to stair-stepping and interpolation techniques used to create uniform sampling rates in time-series sensor signals, which hinder the detection of incipient anomalies in monitored systems.
Innovation Solution
A system that preprocesses time-series sensor signals by classifying stair-stepped signals, performing interpolation, and training an inferential model to replace stair-stepped values with interpolated estimates, using techniques like cubic-spline interpolation and Multivariate State Estimation (MSET) to improve correlation and detect anomalies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If stair-stepping is used to create uniform sampling rates for all time-series signals, then all signals have the same uniform sampling rate, but the cross-correlation with other signals deteriorates and prognostic-surveillance performance worsens
Solution Approach 1:
The patent introduces an inferential model as an intermediary between the raw sensor data and the prognostic-surveillance analysis. This model learns the underlying relationships between multiple sensor signals and generates inferred values that maintain the temporal structure and correlations necessary for reliable anomaly detection, while still providing uniform sampling rates for processing.
Solution Approach 2:
The patent replaces the simple mechanical stair-stepping approach with a sophisticated inferential modeling system that uses machine learning to predict sensor values. This substitution transforms the problem from a simple data replication task to an intelligent prediction task that preserves signal correlations and improves prognostic performance.
2Ease of operation
If interpolation is used to fill in missing values in low sampling-rate sensor signals, then uniform sampling rate is achieved, but cross-correlation with other signals remains poor
Solution Approach 1:
The inferential model serves multiple functions simultaneously: it fills in missing values to achieve uniform sampling rates, preserves cross-correlations between signals, and provides denoised predictions for prognostic analysis. This multi-functionality resolves the contradiction by making a single system that handles all these requirements.
Solution Approach 2:
The patent changes the fundamental parameter of how missing values are generated - instead of using simple interpolation formulas, it uses learned parameters from training data that capture the true relationships between sensors. This parameter change from deterministic interpolation to probabilistic inference preserves information and correlations.
3Device complexity
If stair-stepped values are used in time-series signals, then data processing is simplified, but measurement precision and anomaly detection capability deteriorate
Solution Approach 1:
The patent performs preliminary action by training the inferential model on historical data before actual prognostic-surveillance operations. This pre-training phase captures the relationships and patterns in the data, enabling the model to make accurate predictions during operation without requiring complex real-time processing, thus resolving the contradiction between simplicity and precision.
Data Source
AI summary
During operation, the system obtains the time-series sensor signals, which were gathered from sensors in a monitored system. Next, the system classifies the time-series sensor signals into stair-stepped signals and un-stair-stepped signals. The system then replaces stair-stepped values in the stair-stepped signals with interpolated values determined from un-stair-stepped values in the stair-stepped signals. Next, the system divides the time-series sensor data into a training set and an estimation set. The system then trains an inferential model on the training set, and uses the trained inferential model to replace interpolated values in the estimation set with inferential estimates. Next, the system switches roles of the training and estimation sets to produce a new training set and a new estimation set. The system then trains the inferential model on the new training set, and uses the trained inferential model to replace interpolated values in the new estimation set with inferential estimates.