A study of impact assessment of utility facility enquiry for road excavation work

Abstract

The Mark Plant Circulation (MPC) of road excavation work is a process dealing with work requests in obtaining underground utility facilities information of the utilities and government agencies. Before making a reply from a facility owner to the work requesters, an impact assessment is carried out to avoid damage, to protect facilities and to minimize the public inconvenience. In case of emergency, an immediate assessment has to be carried out and assistance has to be offered on site within two hours. In normal circumstances, the assessment process takes 3 working days. Nowadays, the computerized MPC system has facilitated the data exchange of underground utility facility among utility undertakings by electronic means. However, the utility undertakings still require extensive efforts in assessing the MPC's impact. The difficulties arise mainly due to too many enquiries and variables involved, no rules for making the assessment and, no guidelines for the separation of the impact significance. The process of assessing impact is very time consuming and heavily based on the experience of the assessors. Inconsistent decisions and recommendations may be drawn by different assessors. Clearly, there is an urgent need for the research overcoming those problems. This research has tackled the problems by four investigation phases in cooperation with 11 experiments of data mining. The databases for the investigations cover the MPC project data of time variance and non-time variance, and the spatial position attributes. The experiments assessed the impact of the data from the enquiries and classified them into five distinct categories: High-Plus, High, Medium, Low and None. Phase one examined the influence of variables with and without time variance. Two experiments were conducted using the decision tree and rule sets approaches. It was found that excluding the variables of time variance gave a better result of prediction accuracy. Phase two dealt with the data transformation for better accuracy. The variables of time variance converted to the time independence period, and normalized the unstructured text data. Phase three searched for the region in developing the assessment rules. A set of variables based on the spatial relationship of the work location with the facilities (man-holes and duct routes) built underneath of the road was established. The prediction accuracy was greatly improved. Phase four examined the assessment model by combining variables found in Phases two and three. The classification model for impact has been established. It contains 8 desirable variables, 37 rules that can assess an enquiry to a class of 5 distinct impact categories with the accuracy of 97%. Five modules were developed on platform of the computerized MPC system. It formed as a test rig. The test rig was developed to compare the performance of the Artificial Intelligence (AI) based classification model with Manual assessment approaches. The purpose of the modules is to ensure that the assessing processes are operated correctly for the comparison of assessment's efficiency and effectiveness. To ensure the comparison is fair and consistent, an expert from a utility undertaking with 5 years' experience in the field was appointed. The impact was marked and recorded in a time sheet in a regular basis for comparison. With the AI assessment approach, three developed modules for data extraction, spatial data interpolation & extraction, and prediction analysis were connected to the classification model as the AI assessing packages. The impact assessment process was in batch mode and could handle the MPC enquiries automatically. For the Manual assessment approach, two interfaces, Web-based and Geographical Information System (GIS) based were developed to provide the means for the expert accessing the MPC data as fed to inputs of the AI assessing packages. The inputs of both approaches were obtained from the identical databases of MPC. Comparisons were drawn. The AI assessment approach utilized less resources and time taken. The accuracy of the assessment given by the classification model exceeded 97% for an electricity and power company. The exploration of applying the established classification model to other industries in assessing the impacts was also investigated. The results of comparison have proved that the use of the AI assessment approach performed in a more efficient and effective way. The developed test rig can easily be further extended to set up an operational AI based decision support system for a utility undertaking.

Date of Award	17 Feb 2010
Original language	English
Awarding Institution	City University of Hong Kong
Supervisor	Kin Lim John HO (Supervisor)