This paper presents a low-cost and intelligent Inter- net of Things (IoT) system for gas, environmental, and health monitoring in hazardous industrial environments. The proposed system integrates low-power sensing modules with LoRa communication and an Attentional Bi-LSTM model deployed at the edge. Distributed sensor nodes, built on Arduino microcontrollers, collect temperature, humidity, and gas concentration data using DHT and MQ-series sensors. These readings are transmitted via LoRa to a NanoPi Neo Plus2 gateway, where the Attentional Bi-LSTM model performs multivariate time-series forecasting for early hazard prediction by capturing temporal dependencies and dynamically focusing on critical features. Local edge processing eliminates cloud dependency, reduces latency, and enables immediate visual and auditory alerts through reverse LoRa communication in case of potential risks. Experimental evaluation demonstrates stable LoRa transmission exceeding 1 km in obstructed environments at a baud rate of 9600 bps, along with high predictive accuracy and robust system performance. By combining LoRa-based wireless sensing with edge-deployed deep learning, the proposed system provides a scalable, energy- efficient, and practical solution for proactive gas detection and environmental and health monitoring in safety-critical industrial applications.

Keywords