Sensor-Based Device Association Using Time-Series Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In industrial settings, manually assigning devices to groups based on limited attributes is prone to error, and existing systems struggle to efficiently identify related devices using time-series data from sensors due to varying frequencies and noise levels, especially with hundreds or thousands of devices equipped with thousands to millions of sensors.
Innovation Solution
The system employs unsupervised learning techniques to reduce the dimensionality of time-series data, using a temporal autoencoder comprising a Convolutional Neural Network (CNN) and Bi-directional Long Short Term Memory (Bi-LSTM) network to transform data into a low-dimensional latent space, facilitating clustering and identifying related devices based on sensor data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If manual assignment of devices to groups is used based on limited attributes, then the process is simple to operate, but the accuracy and reliability of device grouping deteriorates
Solution Approach 1:
The patent replaces manual mechanical assignment with automated computational analysis. A computing system automatically groups devices by analyzing time-series sensor data, temporal patterns, and contextual information, eliminating manual errors while maintaining operational simplicity through automated processing.
Solution Approach 2:
The patent introduces an intermediary computational layer that processes sensor data between the devices and the grouping system. This intermediary analyzes temporal patterns and relationships in sensor data to automatically determine device relationships, improving grouping accuracy without requiring manual intervention.
2Measurement precision
If device grouping is based on comprehensive sensor data analysis, then the accuracy of identifying related devices improves, but the computational complexity and processing time deteriorates
Solution Approach 1:
The patent segments the comprehensive sensor data processing into distinct analytical components: temporal pattern analysis, contextual information processing, and relationship determination. This segmentation allows each component to be optimized independently, reducing overall computational complexity while maintaining high identification accuracy.
Solution Approach 2:
The patent transforms raw sensor data into meaningful parameters by analyzing temporal patterns and contextual relationships. By changing the parameter representation from raw sensor values to temporal patterns and relationships, the system achieves high accuracy with reduced computational complexity.
3Speed
If traditional device grouping methods are used with limited attributes, then the processing speed is fast, but the adaptability to dynamic device relationships deteriorates
Solution Approach 1:
The patent implements dynamic device grouping by continuously analyzing temporal patterns in sensor data. The system adapts to changing device relationships by detecting temporal pattern changes over time, allowing the grouping structure to dynamically adjust to evolving device relationships while maintaining efficient processing through pattern-based analysis.
Solution Approach 2:
The patent enables continuous monitoring and analysis of sensor data to maintain up-to-date device groupings. By continuously analyzing temporal patterns, the system adapts to dynamic relationships without requiring periodic reprocessing of all data, maintaining both speed and adaptability through continuous useful action.
Data Source
AI summary
A system includes reception of a set of time-series data from each of a plurality of sensors, each of the plurality of sensors associated with one of a plurality of hardware devices, determination of a plurality of clusters based on the sets of time-series data, assignment of each set of time-series data to one of the plurality of clusters, and determination of associations between the plurality of hardware devices based on the assignments of time-series data to clusters.