Classification of Construction Equipment Working Types using GNSS sensor data and Machine learning

Abstract

The size of the AI ​​market and construction equipment market related to the construction industry continues to grow, and research to analyze the activity of construction equipment through machine learning technology is ongoing. The classification of construction equipment working types is considered to play an important role in solving the difficulties of systematic management of construction equipment by domestic construction managers and increasing the productivity of construction site projects. If the working types of construction equipment are classified in real time and monitoring is possible through the control platform, the site manager will be able to manage the idle and operation time of the construction equipment through remote control without being directly put to the site. In addition, if information on the working types of construction equipment (idle and operating time, activity cycle time of construction equipment) and soil transport information (number of soil transports, soil quantity per transport) are calculated, it will be used as feedback data on construction equipment operation. It is thought to be used to improve earthwork productivity. Therefore, in this study, high-precision VRS-RTK GNSS (Global Navigation Satellite System) equipment is installed on construction equipment Dozers and Rollers to acquire location data and to classify construction equipment working types in real time by utilizing the machine learning data, construction equipment work video, and machine learning software. In the process of the classification, we search for improvements in the classification of construction equipment (Dozers and Rollers) working types, and select an appropriate machine learning algorithm. In addition, in order to select the location data acquisition time interval when location data is used in the classification of construction equipment working types through machine learning in the future, the most efficient data acquisition time interval is selected in terms of the amount of data and the classification result by classifying the construction equipment working types for each time interval (1-second~10-seconds).