Abstract

The National Research Council of Canada reported that rehabilitation of municipal water systems between 1997 and 2012 would cost $28 billion (NRC, 2004). With the rapid increase of new installations, the need for replacement and repair of pipe utilities and also the demand for trenchless excavation methods, increase. This must be done with minimum disruption to public. One alternative to reduce disruption is to use horizontal directional drilling (HDD) for new pipe installation scenarios. Consequently, contractors, engineers, and decision makers are facing continuous challenges regarding to estimation of execution time and cost of new pipe installations, while using HDD. This is because productivity prediction and consequently the cost estimation of HDD involves a large number of objective and subjective factors that need to be considered. It is well known that prediction of both productivity and cost is an important process in establishing and employing management strategies for a construction operation. This calls for the need of developing a dedicated HDD productivity model that meets present day requirements of this area of construction industry. There are two main objectives of the current research. The first objective is to identify the factors that affect productivity of HDD operations. The second objective is to develop a productivity prediction model for different soil conditions. To achieve these two objectives a thorough literature review was carried out. Thereafter, data on potential factors on productivity were collected from HDD experts across North America and abroad. Following data collection, the current research identified managerial, mechanical, environmental and pipe physical conditions parameters operating in three types of soils: clay, rock and sandy soils. Prior to model development, Analytical Hierarchy Process (AHP) technique was used to classify and rank these factors according to their relative importance. A neurofuzzy (NF) approach is employed to develop HDD productivity prediction model for pipe installation. The merits of this approach are that it decreases uncertainties in results, addresses non-linear relationships and deals well with imprecise and linguistic data. The following eight factors were finally selected as inputs of the model to be developed: operator/ crew skills, soil type, drilling rig capabilities, machine conditions, unseen buried obstacles, pipe diameter, pipe length and site weather and safety conditions. The model is validated using actual project data. The developed NF model showed average validation percent of 94.7%, 82.3% and 86.7%, for clay, rock and sand, respectively. The model is also used to produce productivity curves (production rate vs. influencing factors) for each soil type. Finally, an automated user-friendly productivity prediction tool (HDD-PP) based on present NF model is developed to predict HDD productivity. This tool is coded in MatLab ® language using the graphical user interface tool (GUI). The tool was used to test a case study. It was proved to be helpful for contractors, consultants and HDD professionals in predicting execution time and to estimate cost of HDD projects during the preconstruction phase in the environment of imprecise and noisy inputs.