Artificial Intelligent in Construction Management

Only available on StudyMode
  • Download(s) : 590
  • Published : April 27, 2011
Open Document
Text Preview
National Taiwan University of Science and Technology

Artificial Intelligence for Project Management
Final Report
Instructor: Prof. Rayson Chou
Student: Bui Quang Nha: M9705809
Fillip Korsa: E9815001

Table of Contents
1.2.Business needs for pavement life cycle prediction5
2.Project statement5
2.1.Problem definition5
2.2.Project objective6
3.Data description and preprocessing8
3.1.Data description8
3.2.Data preprocessing9
4.Data mining techniques12
4.1Linear Regression12
4.2Neural Network13
4.3Support Vector Machine14
4.4K-fold cross validation15
4.5Performance Evaluation16
5.Analysis Process17
5.1Feature selection17
5.2Data for cross validation18
5.3Modeling in SPSS20
5.4Results and evaluation24
Data in details30
International Roughness Index30
Antislip characteristics / Coefficient of friction31
Macrotecture MPD33
Vehicle records35

In this thesis we deal with a problem of pavement lifetime cycle – pavement serviceability modeling which plays the crucial role in Pavement Management System (PMS). To estimate the time when the damage caused in upper layers begins to spread to lower layers represents a serious problem which scientist all over the world aspire to get on done. Further if we can obtain the damage influence caused by every single heavy vehicle on the pavement we can establish reasonable prices for its usage. In this work we are going to use several modeling techniques such as Support Vector Machine (SVM), Linear Regression (LR), also artificial neural networks (ANNs) were applied for pavement modeling to improve decision-making process.

1. Introduction
1.1. Background
In developed countries with completed highway systems, governments have to spend large amount of money every year to ensure high quality of pavement surface which is most important for speed and safety of transportation. It’s impossible to maintain the whole system frequently as the budget will run out soon. Due to many affecting factors, those systems are gradually deteriorated and need to be maintained. However, system of pavement is influenced by huge number of entering variables such as different parameters of vehicles, weather conditions, soil conditions or pavement construction so to predict the serviceability of the pavement becomes very complex. 1.2. Business needs for pavement life cycle prediction If we can predict the suitable time period for replacing upper pavement layers, then we can save a lot of money and resources and provide high quality of pavement surface. Furthermore if we can obtain the damage influence caused by heavy vehicles on pavement we can establish reasonable price for usage.

Highway directorates in developed countries require creation of well organized pavement management system with long –term pavement analysis usable for recovery planning and for tolling vehicles. For this reason it is necessary to develop appropriate pavement performance models, otherwise deferring maintenance would have no technical or economic consequence.

1. Project statement
2.3. Problem definition
In this part we describe in more details which information are required for the business application of our project by introducing our goals. Goals in points:
* Compare several prediction models using different types of prediction approaches with real estate of pavement and make a recommendation; * Compare the influence of every single inputs especially of large heavy vehicles (FHWA Vehicle Classification) and set up price for using the highways in order to keep the high quality level of pavement surface in next year; * Evaluate relation between caused damage and following...
tracking img