Online fault detection and localization of multiple oil pipeline leaks using model-based residual generation and friction identification

This study proposes a model-based fault detection method for identifying and pinpointing simultaneous leaks in a pipeline, utilizing flow and pressure sensors positioned at the ends of the pipeline. Nonlinear system equations were derived using continuity and momentum principles. It explores a scheme that involves online residual generation for detecting and localizing multiple leaks. This approach incorporates a two-stage decision-making algorithm integrating leak localization with real-time friction parameter identification, which is essential for estimating new simultaneous leaks in the pipeline. A nonlinear observer continuously identifies friction parameters, which is necessary due to oil's non-Newtonian behavior at all temperatures. Additionally, it constructs a dynamic model considering equivalence in the steady state of a leak's position with two simultaneous leaks. An identification process is proposed for estimating the specific parameters of the equivalence model by minimizing the quadratic error between pipeline data and steady-state class members. Compared to existing methods, its main innovation is to update oil pipeline friction functions in real time for all simultaneous leak scenarios. OLGA software simulates oil pipeline realistic operation. Results of the leak detector with different scenarios offered estimation errors of less than 0.5% in all the cases for an oil pipeline of 50 [km]. The effectiveness of the proposed method is confirmed by the convergence of system state and precision in the localization of various simultaneous leak scenarios.