Multi-Sensor Time Series Clustering Using Dynamic Programming
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning-based approaches struggle to detect similar behavioral patterns among multiple sensors in large systems with a high number of input parameters, as sensor values often vary in different ranges and have phase differences, making it difficult for domain experts to identify sensor similarity.
Innovation Solution
A pattern recognition-based approach using Dynamic Programming (DP) to identify similar time series patterns across sensors, followed by clustering, which is computationally efficient and scalable.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional machine learning-based approaches are used to analyze sensor data, then feature analysis can be performed, but it becomes difficult to detect sensor similarity in large systems with many input parameters due to varying value ranges and phase differences
Solution Approach 1:
The patent transforms the sensor data representation by changing parameters: instead of comparing raw sensor values directly, it computes derivatives (rate of change) and then angles of change. This parameter transformation normalizes the data, making sensors with different value ranges and phase differences comparable. The angle of change becomes the key parameter for similarity detection, resolving the contradiction between accurate similarity detection and system complexity.
2Measurement precision
If domain experts manually detect sensor similarity in large systems, then accurate pattern recognition may be achieved, but it is not feasible when the number of input parameters is in the order of 10^3
Solution Approach 1:
The patent replaces the mechanical/manual process of domain expert analysis with an automated computational system. The system automatically computes derivatives, angles of change, and similarity metrics for all sensors without human intervention. This substitution eliminates the time loss associated with manual analysis of 10^3 parameters while maintaining or improving detection accuracy through consistent algorithmic application.
Solution Approach 2:
The system performs self-service by automatically detecting sensor similarities without requiring domain expert intervention. The automated computation of angular similarity metrics and clustering algorithms enable the system to identify patterns independently, resolving the contradiction between accurate detection and time consumption for manual analysis.
3Productivity
If sensors with different value ranges and phase differences are compared directly, then raw data comparison can be performed, but accurate similarity detection becomes difficult
Solution Approach 1:
The patent applies preliminary actions to the sensor data before comparison: it first computes the first derivative (rate of change) and then the angle of change for each sensor. This preliminary transformation standardizes the data representation, eliminating the impact of different value ranges and phase differences. By performing these transformations beforehand, the system enables accurate similarity detection while maintaining processing efficiency.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Industries deploy a plethora of sensors that are attached to a system or human being, respectively. Under multi-sensor environment scenarios, there is a need to detect which sensors are behaving similarly within a time span. Sensor values often vary in range of values yet depict similar time series characteristic and sometimes have a phase difference in operation, thus making it impossible to detect such sensor similarity in a large system where the number of input parameters/ sensor observations. Systems and methods of the present disclosure determine similar behavioral pattern between time series data obtained from multiple sensors and cluster the sensors. The system implements a pattern recognition-based approach to find the similarity and then applies a Dynamic Programming-based approach to detect similarity in at least two time series data and cluster the sensors and corresponding time series data into specific cluster(s).