CNN-Based Intelligent LPG detection and Safety Control System using IOT

Abstract

This paper introduces an IoT, based intelligent LPG detection and safety control system powered by a CNN, aimed at increasing the safety of industrial locations using LPG. Manufacturing plants, chemical plants, and processing industries are examples of industrial facilities that utilize LPG for their operations and, hence, expose the environment to the risk of gas leakage and fire incidents. Most of the time, gas sensors and alarm units are the only components used in traditional LPG detection systems, thus these systems can generate false alarms and they hardly make the right decisions. In an attempt to overcome the problems of these systems, the authors implemented a system that integrates sensor, based monitoring and IoT technology with a CNN for accurate and smart LPG leak detection.The system packs an LPG gas sensor, a camera module, an ESP32 based IoT controller, and a CNN based detection model.The gas sensor functions 24/7 and keeps track of the surroundings for any unusual concentration of LPG. In case a gas leak situation is figured out, the camera module will snap the area through live images or videos.Then, the CNN model, which was trained by the pattern of LPG leaks, is utilized to check the images.By utilizing IoT capability, this system is able to send real, time data, warning messages, and proof figures to a remote industrial monitoring server or mobile dashboard. Thus, safety officers and plant managers have the opportunity to monitor the conditions of LPG remotely and can react very fast during the emergency situations. The associating of CNN with IoT results in smart decision, making, uninterrupted monitoring, and fast communication.The system under discussion is intended to be cheap, expandable, and compatible with the tough industrial environment. Really, it is able to work in industries based on LPG, chemical plants, storage, and manufacturing units. For industrial gas safety management, the CNN, Based Intelligent LPG Detection and Safety Control System is delivering a trustworthy and capable solution by integrating deep learning, sensor, based detection, and IoT, enabled safety control. It provides early warning of leaks, greatly reduces the false alarm rate, and overall significantly increases the safety of the industry

Introduction

Liquefied Petroleum Gas (LPG) is widely used in industries because of its high energy efficiency and ease of storage and transportation. However, its highly flammable nature makes leakage a serious safety hazard that can lead to fires, explosions, loss of life, and extensive property damage. Traditional LPG detection systems mainly rely on threshold-based gas sensors and manual inspections, which are prone to false alarms, sensor drift, and delayed responses. To overcome these limitations, the paper proposes an intelligent LPG detection and safety control system that integrates Convolutional Neural Networks (CNNs), Internet of Things (IoT) technology, and automated safety mechanisms for reliable, real-time industrial monitoring.

The proposed system combines gas sensor measurements with CNN-based visual verification to improve leak detection accuracy and minimize false alarms caused by environmental factors such as temperature, humidity, or lighting variations. An ESP32 microcontroller serves as the central controller, processing data from MQ-series gas sensors, an ESP32-CAM camera module, and optional weight sensors. When abnormal gas concentrations are detected, the camera captures images that are analyzed by a trained CNN model to confirm the presence of an LPG leak. Once a leak is verified, the controller automatically activates safety mechanisms such as gas shut-off valves, ventilation systems, and alarm units, while simultaneously transmitting alerts and system status to a remote IoT dashboard for real-time monitoring and emergency response.

The literature review highlights the evolution of LPG leakage detection systems from conventional sensor-based and GSM-enabled monitoring to IoT, machine learning, and deep learning approaches. While earlier systems offered remote monitoring, automatic valve control, or improved gas classification, they generally lacked intelligent visual validation or fully automated safety responses. Recent hybrid systems integrating CNNs, IoT, and gas sensors have demonstrated higher detection accuracy and reduced false alarms, although challenges such as computational complexity and implementation cost remain. These studies establish the need for an integrated intelligent framework that combines sensing, vision, automation, and remote monitoring.

The proposed architecture consists of several interconnected modules: LPG Gas Sensing, Vision-Based Detection, Data Processing and Control, Safety Control, and IoT Communication. The workflow continuously monitors gas concentration, compares sensor readings with safety thresholds, activates image capture when abnormal conditions occur, processes images using a CNN model, combines sensor and visual outputs for final decision-making, and triggers automatic safety actions if a leak is confirmed. The system also logs all events and provides remote access through a web or mobile dashboard for continuous industrial safety management.

Experimental evaluation under simulated industrial conditions demonstrated that the hybrid CNN-IoT approach significantly improves detection accuracy, reduces false alarms, and shortens response time compared with conventional sensor-only systems. During confirmed leakage events, the system successfully activated gas shut-off valves, ventilation systems, alarms, and IoT notifications in real time. Overall, the proposed solution provides a low-cost, portable, energy-efficient, and intelligent industrial safety system that enhances reliability through multimodal detection, automated emergency control, and continuous remote monitoring, making it well suited for modern LPG-based industrial environments.

Conclusion

In many industrial applications, LPG is consumed in large quantities, so providing greater safety through an intelligent LPG detection and safety Control System using CNN technology has practical applications. The CNN system collects data from different types of gas sensors located in various locations, while CNN provides visual analysis capabilities that can detect the presence of gas. This innovative development solves the problems associated with traditional sensor-only detection methods and supports increased accuracy in detecting real threats from false alarms. An ESP32 Controller is used to carry out real-time data processing, perform automated safety control operations, and provide connectivity to the Internet of Things (IoT) for remote monitoring of systems associated with these processes Experimental data obtained during testing show that the proposed intelligent system correctly detects the presence of LPG leaks, but that it also responds quickly (within 2 seconds) to critical events and initiates safety mechanisms (gas shutoff, ventilation, and alarms) in a timely manner. Alerting users through IoT-based alerts, such as text messages, along with data logging, will provide users with greater situational awareness and support incident investigation.The proposed system has a low cost, is scalable and will be suitable for many industrial environments. The incorporation of deep learning analytics, sensor-based monitoring and IoT technology will enhance operational safety, protect human life and reduce additional risk of industrial accidents or damage to infrastructure.

References

[1] Krizhevsky, A., Sutskever, I., & Hinton, G. E., “ImageNet Classification with Deep Convolutional Neural Networks,” Advances in Neural Information Processing Systems (NIPS), Vol. 25, pp. 1097–1105, 2012. [2] LeCun, Y., Bengio, Y., & Hinton, G.,“Deep Learning,” Nature, Vol. 521, No. 7553, pp. 436–444, 2015. [3] Redmon, J., Divvala, S., Girshick, R., & Farhadi, A.,“You Only Look Once: Unified, Real-Time Object Detection,” Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 779–788, 2016. [4] Howard, A. G., Zhu, M., Chen, B., et al.,“MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications,” arXiv preprint arXiv:1704.04861, 2017. [5] Zhang, Y., Wang, L., & Wang, X.,“Deep Learning-Based Fire Detection Using Surveillance Images,” IEEE Access, Vol. 7, pp. 181219–181229, 2019. [6] Muhammad, K., Ahmad, J., Baik, S. W.,“Efficient Fire Detection for Uncertain Surveillance Environment Using Deep Learning,” IEEE Transactions on Industrial Informatics, Vol. 15, No. 5, pp. 3113–3122, 2018. [7] Chen, T., Zhang, H., & Wang, Y.,“Gas Leakage Detection System Based on Internet of Things,” IEEE Sensors Journal, Vol. 20, No. 12, pp. 6785–6792, 2020. [8] Patil, S., & Kulkarni, P.,“IoT-Based LPG Gas Leakage Detection and Alert System,” International Journal of Engineering Research & Technology (IJERT), Vol. 8, No. 5, pp. 112–116, 2019. [9] Singh, R., Sharma, A., & Verma, P.,“Smart Gas Monitoring System Using IoT,” Proceedings of IEEE International Conference on Smart Systems, pp. 245–250, 2021. [10] Kumar, P., Singh, D., & Kaur, H.,“IoT-Based Smart Gas Leakage Detection System,” International Journal of Advanced Research in Computer Science, Vol. 11, No. 3, pp. 45–50, 2020. [11] Gubbi, J., Buyya, R., Marusic, S., & Palaniswami, M.,“Internet of Things (IoT): A Vision, Architectural Elements, and Future Directions,” Future Generation Computer Systems, Vol. 29, No. 7, pp. 1645–1660, 2013. [12] Espressif Systems,“ESP32 Technical Reference Manual: Wi-Fi and Bluetooth Module,” Version 4.0, 2020. [13] Hanwei Electronics,“MQ-2 Gas Sensor Datasheet for LPG, Smoke and Gas Detection,” 2018. [14] TensorFlow Team,“TensorFlow: Large-Scale Machine Learning on Heterogeneous Systems,” Google Research Documentation, 2021. Available: https://www.tensorflow.org? [15] Bradski, G., & Kaehler, A.,“OpenCV Library: Image Processing and Computer Vision,” OpenCV Documentation, 2020. Available: https://opencv.org?

Copyright

Copyright © 2026 Rajalakshmi S, Santhiya M, Dr. J. Sathya Priya, Mrs. S. P. Revathy. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.