Environmental Sensor Neural Networks for Local Fault Detection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing low-cost environmental sensors, such as electrochemical gas sensors, degrade over time due to poisoning or aging, leading to sensitivity loss and network inaccuracies, necessitating improved fault detection methods that balance reliability, computational efficiency, and data traffic.

Innovation Solution

A sensing device equipped with a measurement module, communication module, and functional analysis module using neural networks to determine temporal features from measurement sequences, combining them with spatial features from neighboring sensors to assess the functional state independently, allowing for localized fault detection and maintenance recommendations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If hardware redundancy is introduced for fault detection, then reliability of fault detection is improved, but device complexity and costs increase

Engineering Contradiction:

Improvefault detection reliabilityVSAvoidhardware complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent uses virtual copies of sensor data from neighboring sensors to create redundant information without physical hardware redundancy. The neural network processes measurements from multiple sensors as if they were redundant copies, achieving fault detection reliability through data replication rather than hardware replication.

Inventive Principle:

Principle #26Copying

Solution Approach 2:

The patent replaces the mechanical/hardware-based fault detection system with a software/neural network-based system. Instead of using additional physical sensors or redundant hardware components, the solution uses computational methods to analyze sensor data and detect faults, substituting mechanical complexity with software intelligence.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Ease of operation

If a mathematical model is used for fault detection, then ease of operation is improved, but measurement precision deteriorates due to model gaps

Engineering Contradiction:

Improvefault detection easeVSAvoidfault detection accuracy

Core Design Contradiction:

Ease of operationVSMeasurement precision

Solution Approach 1:

The patent transforms the fault detection approach by changing the parameters being analyzed. Instead of comparing sensor readings against a fixed mathematical model, the neural network learns dynamic parameters and patterns from actual sensor data, adapting to real-world variations and improving detection accuracy while maintaining ease of operation.

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent implements feedback mechanisms where the neural network continuously learns from sensor data and adjusts its detection criteria. The system uses feedback from multiple sensor readings and temporal patterns to refine fault detection accuracy, closing the gap between theoretical models and actual sensor behavior.

Inventive Principle:

Principle #23Feedback

3Reliability

If centralized server processing is used, then reliability of fault detection is improved, but data traffic and computational power requirements increase

Engineering Contradiction:

Improvefault detection reliabilityVSAvoiddata traffic

Core Design Contradiction:

ReliabilityVSLoss of energy

Solution Approach 1:

The patent segments the fault detection processing from centralized server operations and distributes it to individual sensor nodes. Each sensor performs local neural network inference using its own measurements and those of neighboring sensors, eliminating the need to transmit large amounts of raw data to a centralized server while maintaining detection reliability through distributed intelligence.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent enables sensor nodes to perform self-diagnosis and fault detection independently using local computational resources. The neural network is deployed at the edge devices, allowing them to autonomously detect faults without relying on centralized server processing, thereby reducing data traffic and energy consumption associated with cloud communication.

Inventive Principle:

Principle #25Self-service

Data Source

PatentUS12456366B2Sensing device for sensing an environmental parameter and method for determining information about a functional state of a sensing device

Publication Date: 2025.10.28 INFINEON TECHNOLOGIES AG

AI summary

In accordance with an embodiment, a sensing device for sensing an environmental parameter includes a measurement module configured for providing a sequence of measurement values in dependence on the environmental parameter; a communication module configured for communicating with a further sensing device; and a function analysis module coupled to the measurement module and the communication module. The function analysis module configured for using a neural network for determining a first temporal feature on the basis of the sequence of measurement values, and determining, on the basis of the first temporal feature and on the basis of a second temporal feature provided by the further sensing device, information about a functional state of the measurement module.