Unlabeled Sensor Clustering for Real-Time Device Failure Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional machine learning approaches fail to detect device failures in real-time due to the absence of annotated data and supporting information in industrial IoT networks, making it challenging to identify abnormal behavior in devices as they degrade.
Innovation Solution
A processor-implemented method that computes key performance indicators from unlabeled sensor observations, applies a windowing technique, optimizes deep learning-based auto-encoder hyperparameters, and uses dimensionality reduction and clustering techniques to predict the operating status of devices, including success, failure, or transition states, and generates alerts accordingly.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional machine learning approaches are used for device failure detection, then the system is simple to implement, but it fails to detect abnormality due to absence of annotated data
Solution Approach 1:
The system performs self-service by using unlabeled sensor data to train itself through unsupervised learning. The autoencoder learns normal device behavior patterns without requiring annotated failure data, and detects anomalies by identifying deviations from these learned patterns. This resolves the contradiction by enabling reliable failure detection despite the absence of annotated training data.
Solution Approach 2:
The system changes the parameter of data labeling from labeled to unlabeled, and transforms the detection approach from supervised to unsupervised learning. By computing KPIs from unlabeled sensor observations and using dimensionality reduction followed by clustering, the system adapts to work without annotated data while maintaining detection reliability.
2Measurement precision
If deep learning-based auto-encoder with dimensionality reduction and clustering is applied, then abnormal behavior detection accuracy is improved, but computational complexity increases
Solution Approach 1:
The system extracts only the essential features from high-dimensional sensor data through dimensionality reduction techniques. By identifying and removing redundant dimensions while retaining critical information, the system achieves accurate anomaly detection with reduced computational complexity. The extraction of key performance indicators and essential features resolves the contradiction between precision and complexity.
Solution Approach 2:
The system transforms the data from high-dimensional sensor space to lower-dimensional embedding space through dimensionality reduction, then applies clustering in this transformed space. This dimensional transformation maintains detection accuracy while significantly reducing computational complexity by operating in a compressed feature space rather than the original high-dimensional space.
Data Source
AI summary
Conventionally, detecting time when a device is going to fail in real time has been a real challenge given the associated constraints and requirements. Due to absence in any supporting information or annotated data, traditional approaches have failed to detection abnormality in devices. Present disclosure provide systems and methods for detecting abnormal behaviour of a device from associated unlabeled sensor observations wherein KPIs are computed based on unlabeled sensor observations of at least two sensor parameters and windowing technique is applied on modified dataset to obtain windowed dataset based on which hyper-parameters of deep learning-based auto-encoder are optimized to obtain set of embeddings. Dimensionality reduction technique is applied on the embeddings to obtain embeddings with reduced dimension. Density based clustering technique with hyper-parameters is applied on embeddings with reduced dimension and cluster(s) for unlabeled sensor observations are obtained. Cardinality is assigned to cluster(s) to predict abnormal behaviour of the device.