Deep Learning Anomaly Detection for Ultra-High Sensitivity Gas Signal Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional gas detection technologies face challenges in detecting target signals with very low concentrations, as they are limited by the signal-to-noise ratio and the concept of limit of detection (LOD), which makes it difficult to distinguish micro-scale signals from noise.
Innovation Solution
The method employs deep learning-based anomaly detection by analyzing noise signals from sensors using an artificial neural network. The network is trained with normal noise signals, and the noise signals from sensors containing target substances below the LOD are input to determine the presence of the target signal.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional sensor detection methods are used, then the sensor can detect signals above the limit of detection (LOD), but it cannot distinguish micro-scale signals from noise when the signal-to-noise ratio is low
Solution Approach 1:
The patent introduces an artificial neural network as an intermediary between the sensor output and the detection decision. The neural network processes the raw sensor signals, learning to distinguish between noise and actual target signals through training data. This intermediary system enables the detection of micro-scale signals that would otherwise be indistinguishable from noise, effectively resolving the contradiction between detection sensitivity and signal distinction difficulty
Solution Approach 2:
The patent transforms the detection approach by changing from direct threshold-based signal detection to neural network-based pattern recognition. By training the neural network with various signal and noise patterns, the system learns optimal detection parameters and decision boundaries, enabling accurate detection even when traditional signal-to-noise ratio thresholds would fail
2Reliability
If the limit of detection (LOD) concept with signal-to-noise ratio of 3 or less is used, then detection can be performed with statistical confidence, but it is limited in determining micro-scale signals equal to or less than LOD
Solution Approach 1:
The patent applies preliminary action by training the artificial neural network in advance with extensive datasets containing both noise-only samples and samples with target signals at various concentrations. This pre-training process enables the network to learn the statistical characteristics of noise and the subtle patterns of micro-scale signals, allowing it to make reliable detection decisions even for signals at or below the conventional LOD with higher confidence than traditional methods
3Measurement precision
If sensor sensitivity is increased by adjusting nanosubstances or surface chemical structure, then the sensor can detect very small amounts of gas, but reproducibility and universality decrease and processing cost increases
Solution Approach 1:
The patent substitutes the complex physical-chemical optimization of sensor materials with a computational approach. Instead of adjusting nanosubstances and surface chemical structures to enhance sensitivity, the system uses an artificial neural network to process and interpret signals from standard sensors. This replacement maintains sensor reproducibility and universality while achieving ultra-high detection sensitivity through software-based signal analysis rather than complex material engineering
Data Source
AI summary
Disclosed are an ultra-high sensitivity target signal detection method based on analysis of a noise signal of a sensor using deep learning based anomaly detection and a system using the same. More particularly, disclosed are a method and apparatus for receiving a noise signal from a sensor and inputting data to an artificial neural network trained with a normal noise signal to determine whether or not a target signal is present. The target signal detection method is capable of detecting a target signal having a very low concentration that can be detected by a conventional sensor, whereby the target signal detection method is useful in developing an ultra-high sensitivity sensor.