Abstract
This thesis explores the buckling behaviour of cylindrical oil storage tanks used in power generation, with a focus on the frequent failures reported in Eskom facilities in South Africa. These tanks play a vital role in maintaining the operational efficiency of power plants, yet their susceptibility to buckling under diverse loading conditions often results in significant downtime, environmental hazards, and operational inefficiencies. In a bid to address this frequent and unplanned failure, the research employs Finite Element Analysis (FEA) and Machine Learning (ML) to address these challenges, by offering advanced solutions to understand, predict, and mitigate buckling phenomena.
Finite Element Analysis was utilised to simulate the effects of hydrostatic pressure, wind loads, and geometric configurations on tank stability, thus providing detailed insights into stress distributions, critical buckling loads, and failure mechanisms. The behaviour of the storage tank under high internal pressure resulting from faulty valves, and during loading and discharging of the stored oil, was evaluated for its potential to enhance resistance to buckling. Complementing these simulations, ML models were trained on the FEA-generated data to predict tank behaviour under untested operational scenarios, and the outcome offers a robust framework for proactive maintenance. Analytical methods were employed in the validation of the FEA results, and stress predictions demonstrated an average error that is less than 10%, thus ensuring their practical relevance and reliability. The ML prediction for both the ANN and ANFIS gave a high correlation coefficient (R²) ranging from 0.98-0.99.
The findings of the study demonstrate that existing tank designs are particularly vulnerable to buckling under high internal pressure due to thin walls. Structural enhancements and the use of advanced materials can significantly improve buckling resistance. Furthermore, predictive maintenance strategies that combine FEA and ML
xi
reduce operational risks and costs by identifying potential failures early, thus extending the service life of tanks and minimising downtimes.
This study not only addresses the specific challenges faced by oil storage tanks but also contributes broadly to the field of structural engineering. Its methodologies and insights have wider implications for enhancing the safety, reliability, and efficiency of critical infrastructure in power generation facilities, combining computational precision with actionable engineering innovations to accurately predict the useful life of the storage tank under diverse operating conditions.