Sensor Time Series Change Point Detection Across Multiple Characteristics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods are inadequate for detecting change points in continuous multi-dimensional time series of sensor values or simultaneously evaluating different characteristics of time series to identify various types of change points, which hinders early anomaly detection in manufacturing processes.
Innovation Solution
A method that divides a time series of sensor values into evaluation windows, analyzes at least two different characteristics within each window, and uses machine learning algorithms to identify change points, optionally combining characteristics into a common one for simplified evaluation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional change point detection methods are used, then simple time series analysis is possible, but multi-dimensional time series and multiple change point types cannot be detected simultaneously
Solution Approach 1:
The patent segments the time series analysis by dividing it into multiple independent evaluation windows, each analyzing different characteristics (e.g., level changes, trend changes, seasonal changes). This segmentation allows simultaneous detection of multiple change point types without requiring a single complex unified model, thereby increasing versatility while managing complexity.
Solution Approach 2:
The patent creates a universal detection framework that can identify multiple types of change points (level changes, trend changes, seasonal changes) using a single integrated system. The evaluation windows are designed to be multi-functional, capable of assessing different characteristics of the time series data simultaneously, thus improving adaptability without proportionally increasing device complexity.
2Reliability
If multiple characteristics of time series are analyzed simultaneously, then comprehensive change point detection is achieved, but resource requirements increase
Solution Approach 1:
The patent divides the time series into separate evaluation windows, each focused on detecting specific types of change points based on particular characteristics. This segmentation allows the system to analyze multiple characteristics comprehensively while processing them in manageable, resource-efficient units rather than requiring simultaneous full-scale analysis of all characteristics across the entire time series.
Solution Approach 2:
The patent applies partial action by analyzing only the most relevant characteristics for each specific evaluation window rather than computing all possible characteristics across the entire dataset. Each evaluation window performs a focused analysis on specific change point types, achieving reliable detection for each characteristic while reducing overall computational resource requirements compared to a exhaustive simultaneous analysis.
Data Source
AI summary
A method for ascertaining at least one change point in a time series of sensor values. The method includes: providing a time series of sensor values and dividing the time series of sensor values into at least one evaluation window; for each of the at least one evaluation window, ascertaining at least two different characteristics of the sensor values contained in the corresponding evaluation window; for each of the at least one evaluation window, ascertaining whether the sensor values contained in the corresponding evaluation window have at least one change point based on the at least two different characteristics of the sensor values contained in the corresponding evaluation window; and providing information about ascertained change points.