| 520 |
|
|
|a The reliability of a buried steel pipeline is strongly influenced by external corrosion due to soil. However, the external corrosion does not affect the pipeline equally at all locations and corrosion does not grow at the same rate throughout a pipeline. Therefore, the need of reliable predictive corrosion model is of great interest among researchers and engineers. The available current models are relying on huge historical data storage; thus, massive excavation works, equipment and technical expertise are required. In addition, the feasibility of these models for practical application in different regions with different climate and soil conditions still remains unknown. Therefore, this research aims to develop a predictive model for underground corrosion with the improved multi-parameter models in which several soil characteristics are considered without the need for a return trip to the field, onsite excavation and the presence of technical expertise. Moreover, the models developed in this study are based on empirical results reflecting a wide range of exposure conditions suitable for Malaysia's site conditions through Component 1 and 2. Two predictive corrosion models based on power law equation were developed using two different approaches at two different locations namely real and simplified sites. The most common applied model used to predict corrosion loss is the power law model (P = ktv), where t is exposure time, and k and v are constant regression of soil parameters. There are a total of 932 mild steel coupons being buried in soil up to 18-month period in 5 different locations and 65 soils samples were analysed for its contents and engineering properties. The results were analysed using statistical methods such as exploratory data (EDA), single linear regression (SLR), principal component analysis (PCA) and multiple linear regression (MLR), while Component 3 was conducted to verify the models using two-way ANOVA (Analysis of variance). From the analysis, the extraction of soil variables related to k and v were successfully obtained. In order to get the best fit of predictive model, the extracted variables are modelled using MLR with 20 combinations of linear equation and embedded in the power law equation. The model revealed that chloride (CL), resistivity (RE), organic (ORG), moisture content (WC) and pH were found to be the most influential variable in predicted mass loss, k while sulphate content (SO), plasticity index (PI) and clay content (CC) appear to be influential with v. The predictive corrosion models based on data from real and simplified sites have yielded reasonable prediction of metal mass loss with R2 score of 0.89 and 0.81 respectively. This research has introduced innovative ways to model the corrosion growth for underground pipeline environment. Moreover, heavy statistical analysis has been utilised to determine the level of influence of soil contents and its engineering properties towards soil corrosivity. The model enables to predict potential metal mass loss, hence the level of soil corrosivity for Malaysia. The knowledge on soil corrosivity may assist pipeline operators in designing effective corrosion mitigation program for their underground assets.
|
