Automated Productivity Models for Earthmoving Operations

Earthmoving operations have significant importance, particularly for civil infrastructure projects. The performance of these operations should be monitored regularly to support timely recognition of undesirable productivity variances. Although productivity assessment occupies high importance in earthmoving operations, it does not provide sufficient information to assist project managers in taking the necessary actions in a timely manner. Assessment only is not capable of identifying problems encountered in these operations and their causes. Many studies recognized conditions and related factors that influence productivity of earthmoving operations. These conditions are mainly project-specific and vary from one project to another. Most of reported work in the literature focused on assessment rather than analysis of productivity. This study presents three integrated models that automate productivity measurement and analysis processes with capabilities to detect different adverse conditions that influence the productivity of earthmoving operations. The models exploit innovations in wireless and remote sensing technologies to provide project managers, contractors, and decision makers with a near-real-time automated productivity measurement and analysis. The developed models account for various uncertainties associated with earthmoving projects. The first model introduces a fuzzy-based standardization for customizing the configuration of onsite data acquisition systems for earthmoving operations. While the second model consists of two interrelated modules. The first is a customized automated data acquisition module, where a variety of sensors, smart boards, and microcontrollers are used to automate the data acquisition process. This module encompasses onsite fixed unit and a set of portable units attached to each truck used in the earthmoving fleet. The fixed unit is a communication gateway (Meshlium®), which has integrated MySQL database with data processing capabilities. Each mobile unit consists of a microcontroller equipped with a smart board that hosts a GPS module as well as a number of sensors such as accelerometer, temperature and humidity sensors, load cell and automated weather station. The second is a productivity measurement and analysis module, which processes and analyzes the data collected automatically in the first module. It automates the analysis process using data mining and machine learning techniques; providing a near-real-time web-based visualized representation of measurement and analysis outcomes. Artificial Neural Network (ANN) was used to model productivity losses due to the existence of different influencing conditions. Laboratory and field work was conducted in the development and validation processes of the developed models. The work encompassed field and scaled laboratory experiments. The laboratory experiments were conducted in an open to sky terrace to allow for a reliable access to GPS satellites. Also, to make a direct connection between the data communication gateway (Meshlium®), initially installed on a PC computer to observe the received data latency. The laboratory experiments unitized 1:24 scaled loader and dumping truck to simulate loading, hauling and dumping operations. The truck was instrumented with the microcontroller equipped with an accelerometer, GPS module, load cell, and soil water content sensor. Thirty simulated earthmoving cycles were conducted using the scaled equipment. The collected data was recorded in a micro secure digital (SD) card in a comma separated value (CSV) format. The field work was carried out in the city of Saint-Laurent, Montreal, Quebec, Canada using a passenger vehicle to mimic the hauling truck operational modes. Fifteen Field simulated earthmoving cycles were performed. In this work two roads with different surface conditions, but of equal length (1150 m) represented the haul and return roads. These two roads were selected to validate the developed road condition analysis algorithm and to study the model’s capability in determining the consequences of adverse road conditions on the haul and return durations and thus on the tuck and fleet productivity. The data collected from the lab experiments and field work was used as input for the developed model. The developed model has shown perfect recognition of the state of truck throughout the fifteen field simulated earthmoving cycles. The developed road condition analysis algorithm has demonstrated an accuracy of 83.3% and 82.6% in recognizing road bumps and potholes, respectively. Also, the results indicated tiny variances in measuring the durations compared with actual durations using time laps displayed on a smart cell telephone; with an average invalidity percentage AIP% of 1.89 % and 1.33% for the joint hauling and return duration and total cycle duration, respectively