The Application of ANN Artificial Neural Network to Pipeline TOLC Metal Loss Database
- Passaworn Silakorn (PTT Exploration and Production PLC) | Taneth Puncreobutr (PTT Exploration and Production PLC) | Thanawin Rakthanmanon (Kasetsart University) | Suchada Punpruk (PTT Exploration and Production PLC) | Chatawut Chanvanichskul (PTT Exploration and Production PLC)
- Document ID
- International Petroleum Technology Conference
- International Petroleum Technology Conference, 14-16 November, Bangkok, Thailand
- Publication Date
- Document Type
- Conference Paper
- 2016. International Petroleum Technology Conference
- 4.2 Pipelines, Flowlines and Risers, 7.6.7 Neural Networks, 7 Management and Information, 7.6 Information Management and Systems, 4.2.3 Materials and Corrosion
- Top of Line Corrosion, ANN, Metal Loss Database, TOLC, Artificial Neural Network
- 4 in the last 30 days
- 103 since 2007
- Show more detail
- View rights & permissions
|SPE Member Price:||USD 7.00|
|SPE Non-Member Price:||USD 23.00|
The major corrosion problem of pipelines in the Gulf of Thailand is TOLC (Top of Line Corrosion). To accurately predict TOLC is very challenging and a current TOLC simulation program cannot generate accurate prediction. ANN (Artificial Neural Network), which was reported to successfully predict TOLC, has been being applied to intra-field pipelines in the Gulf of Thailand of the company with the objective to determine specific TOLC modelling to get more accurate metal loss prediction.
ANN process basically consists of data preparation, network design (architecture), weighted modification (iteration of learning and validation) and result analysis. Parameters for corrosion rate calculation are used as the input of the architecture. RMSE (Root Mean Square Error) calculation is used to measure the accuracy of the model, compared to the real metal loss. The prediction model is finally verified by predicting other pipelines whose data have not been used in the process.
With the application of ANN on 3 pipelines in the first phase, ANN could generate a model that accurately predict TOLC for all of them. Results indicated that RMSEs resulted from ANN were better than those from the simulation program from 2.8 to 6.5 times. The model could also predict quite accurately (visually) the other 2 pipelines which have not been used in the process. Subsequent phase of the project (phase 2) has let ANN learn and develop TOLC prediction model based on 19 data samples of 12 pipelines. With more data, ANN now can generate more accurate TOLC prediction model for company. Also, extended input parameter envelopes expand the limitation of ANN prediction. These would finally allow ANN to accurately predict metal loss for both existing and new pipelines within the same asset.
Finally, the ultimate goal is to reduce pipeline cost. With accurate TOLC prediction and with more confidence in TOLC characteristics, cost of materials can be lessened by the reduction of corrosion allowance of the pipelines.
|File Size||4 MB||Number of Pages||14|
S. Mabbutt, P. Picton, P. Shaw and S. Black. 2012. Review of Artificial Neural Networks (ANN) Applied to Corrosion Monitoring. IOP Publishing. Journal of Physics: Conference Series 364 (2012) 012114. doi:10.1088/1742-6596/364/1/012114.
K. Liao, Q. Yao, X. Wu, W. Jia. 2012. A Numerical Corrosion Rate Prediction Method for Direct Assessment of Wet Gas Gathering Pipelines Internal Corrosion. Energies. Energies 2012, 5, 3892–3907; doi: 10.3390/en5103892.
U. Yolcu, E. Egrioglu, C. H. Aladag. 2012. A New Linear & Nonlinear Artificial Neural Network Model for Time Series Forecasting. Elsevier. doi: 10.1016/j.dss.2012.12.006
M. Khashei, M. Bijari. 2009. An Artificial Neural Network (p, d, q) Model for Time Series Forecasting. Elsevier. doi:10.1016/j.eswa.2009.05.044