NEURAL NETWORK AND MULTIPLE REGRESSION MODELS FOR ESTIMATION OF PRODUCTION RATES IN EXCAVATION OPERATIONS

Emad E. Elbeltagi, Hossam H. Mohamed, Dhaheer A. Thabet

Abstract


Production rate estimation is one of the most frequently discussed topics in construction industry. Production rates of excavation operation in building construction are affected by several factors. Among these factors are: hauling distance, loading area layout, dumping area layout, pile foundation, excavator bucket capacity, size and number of hauling units. Consequently, estimation accuracy here is challenged when the effects of these multiple factors are simultaneously considered. In this paper, a comprehensive review of literature and interview with project managers were performed to identify the most significant factors affecting the production rates excavation operations. Sixteen factors were identified as the most significant factors affect the production rates of such operations. These factors were classified into three categories, namely: 1) Job - Site Conditions, 2) Equipment Characteristics, 3) Management Conditions. The objective in this paper is the development of a suitable tool that can be effectively used to predict the production rates of the excavation operation in building construction projects. For this purpose, field observations were conducted to collect realistic production rates over a period of twelve months (12/7/2009 to 17/7/2010) in the city of Alexandria (Egypt). Eighty-five actual case studies taken from seventeen building projects were used as raw data to develop the proposed neural network (NNM) and multiple regression (MRM) models. These data were randomly divided into two groups: (1) training data (75 actual production rates), (2) validating data (10 actual production rates). In conclusion, comparison between the predictive capabilities of both the best NNM and the best MRM indicates that the NNM outperforms the MRM.

Refbacks

  • There are currently no refbacks.