Optimizing Sensor Placement for Event Detection: A Case Study in Gaseous Chemical Detection.

Area of Science:

• Industrial IoT and Sensor Networks
• Chemical Detection and Monitoring
• Machine Learning for Event Detection

Background:

• Strategic sensor placement is crucial for industrial monitoring and event detection in Industry 4.0 environments.
• Optimal sensor placement research is limited, especially for challenges like gas dispersion in chemical detection.
• Existing methods often lack a balance between detection accuracy and deployment cost.

Purpose of the Study:

• To analyze the impact of sensor placement on event detection accuracy in industrial settings.
• To develop and test effective algorithms for chemical gas detection using optimized sensor configurations.
• To identify sensor placements that maximize detection accuracy while minimizing deployment costs.

Main Methods:

• Utilized deep convolutional neural networks (DCNNs) and decision tree (DT) models for event detection.
• Implemented a non-dominated sorting genetic algorithm II (NSGA-II) for multi-objective optimization of sensor placement.
• Tested detection models on a public dataset of chemical substances collected across five locations.

Main Results:

• The DCNN model achieved 100% accuracy in detecting chemical substances.
• Optimal sensor placement was identified using NSGA-II, balancing accuracy and cost.
• High detection accuracy was achieved using only 30% of the available sensors with refined placement.

Conclusions:

• Strategic sensor placement significantly enhances industrial monitoring and event detection capabilities.
• Combining advanced algorithms like DCNNs with optimization techniques like NSGA-II offers a powerful approach to sensor network design.
• Cost-effective and highly accurate chemical detection systems are achievable through optimized sensor deployment.