Real-time air quality monitoring in Bull Trench Kiln-based Brick industry by calibrating sensor readings and utilizing the Serverless Computing

Abstract

Bull Trench Kiln-based Brick industries are major contributors of Particulate Matter Pollutants (PM2.5 and PM10), which leads to deterioration of air quality, and hence needs to be monitored. As the static air monitoring stations require huge infrastructure and are sparsely located, it is necessary to find alternate ways using recent technologies like IoT, Machine Learning, and Serverless Computing. IoT devices being low cost and portable could be deployed near brick industries, however, they are sensitive to fluctuating ambient air temperature or humidity and hence have to be calibrated and there is a need for powerful back-end infrastructures for storage and analyses of data. The existing approach of transmitting IoT-based data to the cloud is time-consuming, energy inefficient, prone to scaling problems, has poor resource utilization, and is expensive. In this study, the IoT data are calibrated using ML Classifier algorithms and an Automatic Function Triggerer - Function-as-a-Service (AFT-FaaS) method is proposed, which is based on event-driven, Serverless method using AWS Lambda along with the Three-Tier Serverless architecture (Edge Tier, Fog Tire, and Cloud Tier) to deal with the challenges of delay, under-resource allocation, and to reduce the expenditures. The data transmitted to the cloud using the proposed AFT-FaaS method are indexed in ElasticSearch and Kibana is used for the analysis and visualization of data. While performing calibration, Random Forest Classifier was chosen as it had accuracy of above 98%. Finally, an in-depth pricing analysis is performed to better understand the difference between different configurations deployed for Serverless and non-Serverless Computing methods, which shows there is a significant reduction (7 times) of expenses incurred while using Serverless Computing than the non-Serverless Computing methods.