PSI - Issue 54

Vasiliki Panagiotopoulou et al. / Procedia Structural Integrity 54 (2024) 482–489 Vasiliki Panagiotopoulou/ Structural Integrity Procedia 00 (2023) 000 – 000

488

7

The development of a corrosion monitoring SHMP system is guided by three key objectives. First, it aims to verify the feasibility of extending damage-tolerant design principles to address corrosion. This involves determining whether a corrosion pit can be considered equivalent to a standard notch (created by Electrical Discharge Machining) in defining the fatigue endurance limit. Such an evaluation would result in an updated framework for authorizing corrosion during the testing and certification phase, potentially leading to more efficient inspection interval planning in the Maintenance Manual for new designs. Second, the system seeks to assess the criticality of corrosion damage compared to similar mechanical damage. Lastly, it strives to establish correlations between the level of corrosion, its rate of progression, and data from environmental and active sensors. The proposed SHMP system for corrosion monitoring combines experimental data from various corrosion tests and intelligent algorithms capable of learning the relationships between specific damage-sensitive parameters and accurate damage characterization. This involves the utilization of analytical models to enhance corrosion assessment and prediction.

(1)

(2)

Fig. 4 (1) Main Rotor Dampers exposed to marine environment and (2) electrodissolution performed on coupon specimens followed by profilometry.

and

4.1. Model-based approach Corrosion is a process resulting from the interaction between a material and its environment, leading to structural degradation. The modeling of corrosion processes holds significant importance, especially within the aeronautic industry. For each type of corrosion, several models have been developed to predict corrosion rates. One such model, created during the MICAT Project (S. Pintos, N. Queipo, O. T. Rincon, A. Rincon, M. Morcillo, 2000), has demonstrated its validity across various atmospheric conditions and with different metals. This statistical model assumes that corrosion rate varies over time, influenced by various environmental factors specific to each case, such as temperature, humidity, slag corrosion, and more. The probabilistic model specifically addresses pitting corrosion in commonly used metals and alloys in the aeronautic industry. This model provides a mechanistically-based probability model for the nucleation, growth, and coalescence of corrosion pits, and it is incorporated into ML algorithms for the purpose of corrosion damage detection. It is primarily motivated by the environmental effects on exposed surfaces. The objective is to determine the cumulative distribution function for pit sizes, which, in turn, varies with time. As previously mentioned, experimental data collected for corrosive loss serves as valuable prior knowledge used by these models. This data assists in identifying fundamental parameters related to the corrosive process, such as corrosion rate, corrosion depth, and area. This information allows for assessments of the overall residual strength in the affected area. 4.2. Domain adaptation methods Machine learning algorithms, whether supervised or unsupervised, are typically designed assuming that training and test data come from the same distribution. This assumption can lead to poor predictions when applied to diverse case studies, requiring expensive data collection for each structure. In addition to the time and financial concerns,