Developing Real-Time Acoustic Technologies for Enhancing Safety and Efficiency on Construction Job Sites Using Deep Learning Algorithms: Moving towards Automated Construction Site Monitoring

Construction site monitoring is a critical aspect of project management, and it includes enhancing safety and operational efficiency. Acoustic devices could be used to obtain real-time data necessary to monitor construction sites and to provide safety- and efficiency-related information. Traditional methods often fall short in providing timely insights into safety and operational aspects. Thus, this research addresses this imperative by proposing an architecture with the help of machine learning techniques in construction site surveillance. First, a diverse data set of construction site audio recordings was collected to capture a spectrum of activities, machinery, and ambient sounds generally seen on construction sites. Rigorous data preprocessing was then implemented. Subsequently, a deep machine learning convolutional neural network (CNN) model was developed and trained by exploring various architectures and hyper-parameters to enhance accuracy in classifying construction-related audio events. The results showcase the system's effectiveness in accurately identifying construction site sounds, which could result in timely and targeted alerts to stakeholders. Ultimately, this paper holds promise for transforming the way construction sites are observed and managed, thus contributing to a safer and more efficient construction industry.