A Damage Assessment Model, Construction Process Model and Parametric Quantity Model were developed with the purpose of capturing the engineering knowledge involved in the estimating process of bridge repair construction projects. The Damage Assessment Model was used to create a sample database in which detailed inspection data was stored in a format compatible with the existing Pontis?tabase. Detailed inspection data, which provided quantitative values for the different damage types observed in bridges, could be retrieved from the sample database so that data could be used as either input parameters in the knowledge rules that triggered the selection of construction tasks in the Construction Process Model, or data could be used as variables in the equations used to estimate quantities in the Parametric Quantity Model. The Construction Process Model was used to incorporate the logic behind the construction process for different repair methods. The Construction Process Model was composed of seven repair matrices that defined specific repair methods for each Pontis?idge element. Construction tasks were grouped in construction modules that were modeled as flowcharts. Each construction module flowchart was composed of construction tasks arranged in sequential order and decision points that triggered the selection of construction tasks based on input parameters and knowledge rules. Input parameters were provided by the user, retrieved from the model or pre-defined in the model by expert knowledge. The construction modules developed involved construction tasks related to the repair of concrete bridge piles that were damaged due to reinforcement corrosion and related concrete deterioration. Data describing the construction tasks that were considered in the construction module flowcharts were modeled using the entity-relationship model and were stored in the sample database described previously. The Parametric Quantity Model combined data generated by the Damage Assessment Model and the Construction Process Model with additional expert knowledge and parameters into equations that were used to estimate quantities. The author investigated the use of neural networks as a tool to predict actual damage in bridge piles, conducted a preliminary survey to define labor productivity factors and collected data to define the duration of construction activities related to bridge repair.

Preconstruction services play a vital role in ensuring timely approval of infrastructure funds and successful execution of construction projects. Most state DOTs use simple methods such as a percentage of estimated construction costs that has proven to be unreliable. Several studies have developed statistical models using historical data to improve current practices. However, such models have performed poorly, and practitioners have not utilized these models. This study develops and evaluates data mining models such as multiple regression and artificial neural networks and concludes that such models do not provide sufficiently accurate estimates of preconstruction service fees and hours. Subsequently, it proposes an alternative approach using a case-based reasoning (CBR) technique that uses similarity scoring to retrieve the most similar projects. The historical preconstruction service fees and hours of similar projects can be used to estimate preconstruction service fees and hours for a new project and make any adjustment necessary. A spreadsheet tool is developed to implement this CBR technique. The tool provides a simple and flexible platform that enables engineers to extract necessary data and help them in making data-driven estimates. Thus, the tool is expected to aid state DOT engineers in negotiating with consultants with higher confidence.