Multi-gas pollutant detection based on sparrow search algorithm optimized ALSTM-FCN.

Area of Science:

• Environmental Science
• Chemical Engineering
• Computer Science

Background:

• Accurate detection of hazardous gases is crucial for industrial safety and medical diagnostics.
• Existing methods face challenges in reliably identifying and categorizing a wide range of dangerous gases.

Purpose of the Study:

• To propose and evaluate an advanced deep learning model for hazardous gas detection and categorization.
• To compare the performance of the proposed model against established deep learning and conventional machine learning approaches.

Main Methods:

• Development of an attention-based Long Short term memory Full Convolutional network (ALSTM-FCN).
• Optimization of ALSTM-FCN network parameters using the Sparrow Search Algorithm (SSA).
• Evaluation using University of California-Irvine (UCI) datasets and comparison with LSTM, FCN, and various machine learning models (AdaBoost, LR, ET, DT, RF, KNN).

Main Results:

• The ALSTM-FCN model achieved a reliability test accuracy of 99.461%, significantly outperforming LSTM (89.471%) and FCN (96.083%).
• The proposed SSA-optimized ALSTM-FCN demonstrated superior gas categorization accuracy compared to conventional machine learning models.
• SSA optimization proved more effective than PSO, GA, GWO, and CS algorithms for parameter tuning.

Conclusions:

• The ALSTM-FCN hybrid model offers a highly accurate and reliable solution for hazardous gas detection.
• The study highlights the potential of SSA-optimized deep learning for broad-range polluting gas detection applications.