Sliding Window Filter for Trending Data in Mechanical Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current prognosis and fault detection systems in physical systems, such as aircraft, face limitations due to inaccurate data filtering and prediction methods like Kalman filters, which rely on statistical assumptions and lack reliability in predicting future sensor values and detecting faults, leading to potential unexpected failures and false alarms.
Innovation Solution
A trending system utilizing a sliding window filter to calculate upper and lower confidence bounds for data sets, a trend change detection mechanism using convex hulls to identify transition points, and an outlier elimination mechanism to remove statistical outliers, providing a smoothed estimated trend for prognostication and fault detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If Kalman filters are used for data filtering, then the system can implement automated prognosis and fault detection, but the filtering accuracy is insufficient and statistical assumptions may not be valid for some applications
Solution Approach 1:
The patent transforms the filtering problem from a parametric statistical approach (Kalman filters requiring covariance matrices and statistical assumptions) to a non-parametric approach using kernel density estimation. This fundamental parameter change allows the system to adapt to arbitrary data distributions without requiring valid statistical assumptions, thereby simultaneously improving both reliability and measurement precision in fault detection applications.
Solution Approach 2:
The patent replaces the mechanical statistical filtering mechanism with an information-theoretic approach using kernel density estimation and Kullback-Leibler divergence. This substitution eliminates the need for statistical assumptions about data distributions while maintaining the ability to filter and predict sensor data, resolving the contradiction between reliability and measurement precision.
2Extent of automation
If current trending systems are used, then fault detection can be performed, but false detection of future faults occurs due to inaccurate predictions
Solution Approach 1:
The patent implements a feedback mechanism where the kernel density estimation continuously learns from historical sensor data to refine its prediction model. By comparing predicted future values with actual observed values and updating the density estimates accordingly, the system reduces false detections while maintaining automated prognosis capability, thereby improving reliability without sacrificing automation extent.
3Measurement precision
If neural networks and fuzzy logic are used for predictions, then prediction capability is improved, but careful tuning to the problem is required and quantitative estimate of reliability is not provided
Solution Approach 1:
The patent implements a self-service mechanism where the kernel density estimation automatically adapts to the specific problem domain through data-driven learning without requiring manual tuning. The system autonomously learns the underlying data distribution and prediction patterns from historical sensor data, eliminating the complex tuning process required by neural networks and fuzzy logic while maintaining high prediction accuracy.
Solution Approach 2:
The patent introduces kernel density estimation as an intermediary between raw sensor data and fault prediction decisions. This intermediary provides a probabilistic framework that naturally quantifies prediction reliability through density estimates, eliminating the need for complex tuning while providing both accurate predictions and quantitative reliability measures.
Data Source
AI summary
A trending system and method for trending data in a mechanical system is provided. The trending system includes a sliding window filter. The sliding window filter receives a data set of data points generated by the mechanical system. The sliding window filter partitions the data set into a plurality of data windows, and uses the data windows to calculate upper and lower confidence bounds for the data set. Specifically, the sliding window filter calculates an upper confidence bounds and lower confidence bounds for each data point using each of the multiple data windows that includes the data point. The sliding window filter then selects the upper confidence bounds and the lower confidence bounds that results in the smallest mean prediction confidence interval for that data point. This results in a smoothed estimated trend for the data set that can be used for prognostication and fault detection.