Missing Value Imputation for Time-Series Sensor Data Prognostics

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing interpolation techniques for filling missing values in time-series sensor data fail to accurately detect anomalous events, as they treat 'blind spots' as uniform, leading to undetected degradation and failures in prognostic surveillance systems.

Innovation Solution

A two-phase missing value imputation procedure using a multivariate state estimation (MSET) model, which divides data into training and estimation sets, trains on interpolated values, and switches sets to refine estimates, ensuring accurate imputation by leveraging serial and cross-correlation among signals.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If traditional interpolation schemes are used to fill missing values, then the data can be processed continuously, but anomalous events during blind spots cannot be detected

Engineering Contradiction:

Improvedata processing continuityVSAvoidanomaly detection accuracy

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The patent segments the data processing into distinct phases: training phase using interpolated values to build the inferential model, and estimation phase where the model replaces interpolated values with learned patterns. This segmentation allows the system to first ensure data continuity through interpolation, then subsequently improve accuracy by replacing those interpolations with anomaly-aware estimates from the trained model.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent applies preliminary action by first filling missing values with interpolated values to create a complete dataset for training the inferential model. This preliminary completion of data allows the model to learn normal patterns and correlations, which are then used in the estimation phase to accurately reconstruct missing values while preserving anomaly detection capability.

Inventive Principle:

Principle #10Preliminary action

2Device complexity

If simple interpolation is used to fill missing values, then the process is computationally simple, but the imputed values create blind spots that prevent anomaly detection

Engineering Contradiction:

Improveimputation process complexityVSAvoidanomaly detection capability

Core Design Contradiction:

Device complexityVSDifficulty of detecting and measuring

Solution Approach 1:

The patent implements feedback by using the trained inferential model to continuously improve the imputed values. The model learns from the training data where actual values are known, and uses this learned knowledge to provide feedback corrections to the initial interpolated values in the estimation set, thereby eliminating blind spots while maintaining computational efficiency.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The patent changes the parameter of imputation accuracy by transitioning from simple interpolation to model-based inference. The inferential model captures complex temporal and spatial correlations among sensor signals, allowing the system to change from naive value filling to intelligent prediction that preserves anomaly detection capability.

Inventive Principle:

Principle #35Parameter changes

3Measurement precision

If multiple sensors are deployed to monitor critical assets, then monitoring coverage is improved, but missing values increase due to bus saturation and wireless transmission limitations

Engineering Contradiction:

Improvesystem monitoring coverageVSAvoiddata completeness

Core Design Contradiction:

Measurement precisionVSLoss of information

Solution Approach 1:

The patent merges multiple data sources and correction methods by combining the initial interpolated values with corrections from the trained inferential model. The model integrates information from temporal patterns, spatial correlations, and signal dependencies to reconstruct missing data, effectively merging multiple correction approaches into a unified imputation process that restores data completeness.

Inventive Principle:

Principle #5Merging (Combining)

Data Source

PatentUS11775873B2Missing value imputation technique to facilitate prognostic analysis of time-series sensor data

Publication Date: 2023.10.03 ORACLE INT CORP

AI summary

First, the system obtains time-series sensor data. Next, the system identifies missing values in the time-series sensor data, and fills in the missing values through interpolation. The system then divides the time-series sensor data into a training set and an estimation set. Next, the system trains an inferential model on the training set, and uses the inferential model to replace interpolated values in the estimation set with inferential estimates. If there exist interpolated values in the training set, the system switches the training and estimation sets. The system trains a new inferential model on the new training set, and uses the new inferential model to replace interpolated values in the new estimation set with inferential estimates. The system then switches back the training and estimation sets. Finally, the system combines the training and estimation sets to produce preprocessed time-series sensor data, wherein missing values are filled in with imputed values.