- Title
- Deep learning model for demolition waste prediction in a circular economy
- Creator
- Akanbi, Lukman A.; Oyedele, Ahmed O.; Oyedele, Lukumon O.; Salami, Rafiu O.
- Relation
- Journal of Cleaner Production Vol. 274, Issue 20 November 2020, no. 122843
- Publisher Link
- http://dx.doi.org/10.1016/j.jclepro.2020.122843
- Publisher
- Elsevier
- Resource Type
- journal article
- Date
- 2020
- Description
- An essential requirement for a successful circular economy is the continuous use of materials. Planning for building materials reuse at the end-of-life of buildings is usually a difficult task because limited time are usually made available for building removal and materials recovery. In this study, deep learning models were developed for predicting the amount (in tons) of salvage and waste materials that are obtainable from buildings at the end-of-life prior to demolition. Datasets used for deep neural network model developments were extracted from 2280 building demolition records obtained from the practitioners in the UK Demolition Industry. The data was partitioned into training, testing and validation datasets in the ratio 8:1:1. Deep learning models were developed with a deep learning framework in R programming environment. The average R-squared value for the three deep learning models is 0.97 with Mean Absolute Error between 17.93 and 19.04. The models were evaluated with four scenarios of a case study building design. The results of the evaluation show that, given basic features of buildings, it is possible to predict with a high level of accuracy, the amount of materials that would be recovered from a building after demolition. The models developed will provide decision support functionalities to demolition engineers and waste management planners during the pre-demolition audit exercise.
- Subject
- deep learning; deep neural network; buildings' end-of-life; circular economy; building materials; SDG 12; Sustainable Development Goals
- Identifier
- http://hdl.handle.net/1959.13/1432650
- Identifier
- uon:39084
- Identifier
- ISSN:0959-6526
- Language
- eng
- Reviewed
- Hits: 1173
- Visitors: 1165
- Downloads: 0
Thumbnail | File | Description | Size | Format |
---|