Abstracts – Browse Results
Click on the titles below to expand the information about each abstract.
Viewing 1 results ...
Graham, D (2005) Stochastic modelling and analysis of construction processes, Unpublished PhD Thesis, School of Engineering, University of Edinburgh.
- Type: Thesis
- Keywords: artificial intelligence; case-based reasoning; complexity; modelling; planning; prototype development; stochastic
- ISBN/ISSN:
- URL: https://era.ed.ac.uk/handle/1842/12054
- Abstract:
Construction projects frequently overrun and finish over budget. This is, in part, due to the lack of control that construction practitioners - planners and managers - have over the construction schedule at a process level. Contemporary planning tools, such as critical path method (CPM) are adequate for planning at a project level but are not of sufficient detail to help a practitioner allocate resources to a process to maximise the performance. Thus, the aim of this project was to develop a practical computer-based model to enable practitioners to plan projects at a process level and to improve their projects. This research has focused upon a specific type of construction process - those that are stochastic and cyclical - because such processes are: difficult to predict and hence control; and they are widespread throughout construction projects. Examples of processes that are stochastic and cyclical are crane operations, formwork erection and scaffold erection. Initially, a focus was placed on the process of ready-mixed concrete (RMC) supply. A significant amount of data was available for this process, collected through the observations of a previous PhD student. This data was rigorously analysed to: determine if the process requires a non-linear stochastic solution, which it does; determine if the data was valid - a method of identifying outliers in a simple manner was devised to do this; determine the key variables for use in modelling. Guided by this statistical analysis, a discrete event simulation (DES) model of the RMC supply process was developed and validated. This model could not provide accurate estimates of the process, due to the need for a user define the probability distributions that represent the process. This is a complex requirement for a construction practitioner and the issue became known as the complexity problem and a solution was sought in the field of artificial intelligence. Case-based reasoning (CBR) provides solutions to new problems using knowledge of past ones - exactly the role of a practitioner in the above simulation model. Thus, CBR was investigated as a solution to the complexity problem. CBR, when based upon a novel knowledge retrieval method, proved to be capable of solving the complexity problem. A hybrid model, CBRSim, was developed to fulfil the original aim of this thesis. In this model, CBR is used to select the probability distributions to represent the process, given a set of user-defined operating conditions. These probability distributions are then supplied to a discrete-event simulation model that accurately recreates the process under the user-defined operating conditions. CBRSim was validated and found to be able to predict the productivity of the RMC supply process (volume delivered per hour) to within +7- 3% accuracy. Experimentation with various aspects of the process, such as resource allocation or supplier selection is performed to help illustrate the practical nature of the model. CBRSim was then applied to another stochastic and cyclical process: earthmoving. CBRSim was found to be more accurate in modelling earthmoving than RMC supply, indicating a generic modelling capability.