IoT Sensor Period Filtering for Reliable Anomaly Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing anomaly detection systems in IoT networks face challenges with manual systems' high manpower and time costs, and automatic systems struggle with feature extraction limitations and reliability due to the mix of normal and anomaly data, especially in unstable environments.
Innovation Solution
A statistical-based anomaly period detection system integrating a statistical analysis technique and an unsupervised learning-based neural network model, specifically using an AutoEncoder, to automate labeling and improve detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual anomaly detection systems are used, then detection accuracy can be maintained through human judgment, but manpower and time costs increase significantly
Solution Approach 1:
The system performs preliminary actions by automatically labeling sensor data as normal or anomaly periods using statistical analysis and AutoEncoder models before human review. This preprocessing step filters out obvious anomalies, allowing human operators to focus only on complex cases, thereby reducing time cost while maintaining detection accuracy.
Solution Approach 2:
The patent introduces an intermediary automated detection system that bridges manual and fully automatic approaches. The statistical-based period filtering model and AutoEncoder act as intermediaries to handle routine detection tasks, freeing human operators from repetitive work while preserving human judgment for challenging cases.
2Loss of time
If automatic anomaly detection systems are used, then manpower costs are reduced, but detection reliability decreases due to feature extraction limitations and data mixing
Solution Approach 1:
The patent segments the detection process into distinct phases: statistical analysis for initial filtering, AutoEncoder-based deep learning for pattern recognition, and human review for final verification. This segmentation allows each component to specialize in specific tasks, improving overall reliability while maintaining automation efficiency.
Solution Approach 2:
The system applies local quality by using different detection methods for different data characteristics. Statistical analysis handles routine patterns, while AutoEncoder models focus on complex anomalies. Human operators concentrate on edge cases, ensuring each segment receives appropriate attention quality.
3Productivity
If statistical analysis techniques are used alone, then processing speed is maintained, but detection accuracy suffers from inability to handle complex patterns
Solution Approach 1:
The patent merges statistical analysis with deep learning-based AutoEncoder models to combine the speed advantages of statistical methods with the pattern recognition capabilities of neural networks. This hybrid approach maintains processing speed while significantly improving detection accuracy for complex anomalies.
Solution Approach 2:
The detection system functions as a composite methodology, combining traditional statistical techniques with modern deep learning approaches. Each component contributes its strengths: statistical analysis provides computational efficiency and the AutoEncoder models provide sophisticated pattern recognition, creating a detection system superior to either method alone.
4Measurement precision
If AI models are used alone, then complex pattern recognition improves, but computational complexity and resource requirements increase
Solution Approach 1:
The system performs preliminary statistical analysis to filter and preprocess data before feeding it to the AutoEncoder model. This preliminary action reduces the complexity of input data, allowing the AI model to focus on complex pattern recognition without being overwhelmed by raw data volume, thereby reducing computational complexity while maintaining accuracy.
Data Source
AI summary
Proposed is a server device that supports creation of a statistical-based period filtering model. The server device may create a smoothing period list by performing a preprocessing process on sensor data received from a sensor, based on a pre-stored first parameter, and smoothing the sensor data based on the first parameter. The server device may also create a direction period list by defining directionality of the sensor data based on a predetermined reference value. The server device may further classify the smoothing period list and the direction period list into a normal period pattern or an anomaly period pattern by performing a filtering process on the smoothing period list and the direction period list based on a pre-stored second parameter.