Pile Construction Productivity Assessment Tarek M. Zayed 1 and Daniel W. Halpin 2 Abstract: Bored piles are vital elements for highway bridge foundation. A large number of factors oversees productivity and cost estimation processes for piles, which creates many problems for the time and cost estimators of such process. Therefore, current study is designed to diagnose these problems and assess productivity, cycle time, and cost for pile construction using the artificial neural network ANN. Data were collected for this study through designated questionnaires, site interviews, and telephone calls to experts in different construction companies. Many variables have been considered to manage the piling construction process. Three-layer, feed forward, and fully connected ANNs were trained with an architecture of seven input neurons, five output neurons, and different hidden layer neurons. The ANN models were validated and proved their robustness in output assessments. Three sets of charts have been developed to assess productivity, cycle time, and cost. This research is relevant to both industry practitioners and researchers. It provides sets of charts for practitioners’ usage to schedule and price out pile construction projects. In addition, it provides researchers with a methodology of applying ANN to pile construction process, its limitation, and future suggestions. DOI: 10.1061/ASCE0733-93642005131:6705 CE Database subject headings: Pile foundations; Bored piles; Construction; Productivity; Costs; Neural networks. Introduction Bored piles drilled shaft are widely used in the foundation of highway bridges nowadays. There are several types of problems that control the installation of bored pile foundations, such as subsurface obstacles, lack of contractor experience, and site plan- ning difficulties. An explanation of the effect of the above prob- lems on productivity can be summarized in the following state- ments. The site preinvestigation usually consists of insufficient statistical samples around the foundation area that do not cover the entire area. Soil types differ from site to site or within a site due to cohesion or stiffness, natural obstacles, and subsurface infrastructure construction obstacles. Lack of experience in ad- justing the pile axis, length, and size present a further complica- tion. Piling machine mechanical and drilling problems must be considered. Problems due to site restrictions and disposal of ex- cavated spoil have great effect on productivity. The rate of steel installation and pouring concrete is impacted by the experience of rebar crew and method of pouring. All these problems, no doubt, greatly affect the production of concrete piles on site. In addition there is a lack of research in this field. Because of the above- mentioned problems and others, it is difficult for the estimator to evaluate the piling process productivity. There is a lack of re- search in studying pile foundation construction. Therefore, it is necessary to use sophisticated techniques to analyze the problem and determine the closest optimal solution. The objectives of cur- rent research are: 1 studying the factors that affect productivity and cost of the piling process and 2 applying the artificial neural network ANN to assess pile construction productivity, cycle time, and cost. Pile Installation Productivity Factors Based on studies of the construction process, site interviews, tele- phone calls, questionnaires, and literature review, the following bored pile construction productivity factors were identified Peu- rifoy et al. 1996: • soil type i.e., sand, clay, stiff clay, …etc; • drill type. e.g., auger, bucket; • method of spoil removal, size of hauling units, and space con- siderations at the construction site; • pile axis adjustment; • equipment operator efficiency; • weather conditions; • concrete pouring method and efficiency; • waiting time for other operations i.e., pile axis adjustment; • job and management conditions; and • cycle time. General Concepts of Artificial Neural Networks The artificial neural network consists of a large number of artifi- cial neurons that are arranged into a sequence of layers with ran- dom connections between the layers Tsoukalas and Uhrig 1997. The ANN processing elements e.g., neurons are arranged in lay- ers so that the connections are systematic and the network can be 1 Assistant Professor, Dept. of Building, Civil, and Environment Engineering, Concordia Univ., 1257 Guy St., Montreal, QC Canada, H3G 1M7. 2 Head, Division of Construction Engineering and Management, School of Civil Engineering, Purdue Univ., West Lafayette, IN 47907-1294. Note. Discussion open until November 1, 2005. Separate discussions must be submitted for individual papers. To extend the closing date by one month, a written request must be filed with the ASCE Managing Editor. The manuscript for this paper was submitted for review and pos- sible publication on October 7, 2003; approved on December 23, 2004. This paper is part of the Journal of Construction Engineering and Man- agement, Vol. 131, No. 6, June 1, 2005. ©ASCE, ISSN 0733-9364/2005/ 6-705–714/$25.00. JOURNAL OF CONSTRUCTION ENGINEERING AND MANAGEMENT © ASCE / JUNE 2005 / 705