Environmental transport processes of slow granular filtration using machine learning
In current practice, maintenance of adsorber is only done either when the sensor reaches its limit or by regular maintenance. However, these approaches are considered as passive and might result in huge wastage of resources. Predictive maintenance allows for a more proactive approach. That said, the...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project |
| Language: | English |
| Published: |
2019
|
| Subjects: | |
| Online Access: | http://hdl.handle.net/10356/77443 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | In current practice, maintenance of adsorber is only done either when the sensor reaches its limit or by regular maintenance. However, these approaches are considered as passive and might result in huge wastage of resources. Predictive maintenance allows for a more proactive approach. That said, there are no available tools for predictive maintenance. The main objective of this report is to develop a predictive model to determine the breakthrough behaviour of adsorber. The predictive model involves training a certain period of the initial dataset to predict the subsequent dataset. This allows users to determine the breakthrough time and thus able to plan for maintenance. In addition, the predictive model would also allow for the pre-designing of new adsorbers as parameters of the model can be adjusted to tailor for the needs of new adsorbers. Predictive maintenance was studied with experiments conducted in a lab scale adsorber containing GAC. With varying flowrates, the influent was pumped into the filter and the concentration of influent and effluent were tested with TOC analyser. The results, together with datasets obtained from other literatures, were then utilised in the development of the predictive maintenance model. Root mean squared error between the actual and predicted values was computed to verify the accuracy of the model. It was found that the model was able to successfully predict up to 10% accuracy. However, it did not work for datasets that had already attained breakthrough in the training period. For the model to be within 10% accuracy, a training period of at least 35% is required. Furthermore, predicting datasets with limited datapoints had also proved to be ineffective. |
|---|